@sellable/install 0.1.155 → 0.1.158

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
package/README.md CHANGED
@@ -58,11 +58,11 @@ Auth is stored once at:
58
58
  Claude Code and Codex are configured to launch the same packaged MCP server. The
59
59
  installer also writes Sellable agent definitions from the packaged `agents/`
60
60
  registry, but normal create-campaign runs use only the Message Drafting
61
- background agent. Source discovery and filter/rubric setup stay in the parent
62
- thread with product-native MCP tools unless the user explicitly asks for a
63
- source-comparison/debug run. `get_post_find_leads_scout_registry` returns the
64
- Message Drafting worker for the normal path; `get_source_scout_registry` is
65
- metadata for explicit debug/comparison use.
61
+ background agent (`post-find-leads-message-scout`). Source discovery and
62
+ filter/rubric setup stay in the parent thread with product-native MCP tools.
63
+ `get_post_find_leads_scout_registry` returns the Message Drafting worker for the
64
+ normal path; `get_source_scout_registry` intentionally returns no custom
65
+ source-scout agents.
66
66
 
67
67
  ## Names
68
68
 
@@ -17,20 +17,19 @@ campaign state. The main thread owns approval and campaign writes.
17
17
 
18
18
  ## Source Of Truth
19
19
 
20
- Use the live campaign inputs supplied by the parent thread:
20
+ Use the lean handoff supplied by the parent thread:
21
21
 
22
22
  - `campaignId`
23
- - campaign revision or `campaignUpdatedAt`
24
- - campaign brief summary from the parent, then the current `campaignBrief` /
25
- campaign brief content loaded through Sellable tools
26
- - selected source decision and provider state
27
- - `selectedLeadListId` or selected source list context
28
23
  - `workflowTableId`
29
- - compact initial campaign-table execution-slice basis: copied source row count,
30
- review-batch row count, and review-batch basis/hash. Do not require the
31
- parent to paste a long review-batch row-id list.
32
- - filter basis at branch start: `pending`, `use-filters`, or `skip-filters`
33
- - any already-saved fit/rubric result summaries supplied by the parent
24
+ - concise brief summary: offer, buyer, source context, safe proof, blocked claims
25
+ - concise source summary and source-use rule
26
+ - 3-5 sample workflow-table rows with `rowId`, name, title, company, and short
27
+ signal
28
+ - optional `campaignName`, `selectedLeadListId`, and filter choice
29
+
30
+ Do not require campaign revision, brief hash, copied row count, review-batch
31
+ count, row hash, or a long review-batch row-id list. The branch can verify the
32
+ current state with live Sellable tools.
34
33
 
35
34
  Do not require or hunt for `brief.md`, `lead-review.md`, or `lead-sample.json`.
36
35
  Those files are optional debug context only when the parent explicitly provides
@@ -40,34 +39,49 @@ read stale local markdown files to reconstruct campaign state.
40
39
  All live reads must come from scoped MCP/product tools by campaign and
41
40
  workspace, such as `get_campaign`, `get_campaign_context`, and
42
41
  `get_rows_minimal({ tableId: workflowTableId })`, or from equivalent parent
43
- thread payloads. Load the current campaign brief and the current review-batch
44
- rows from those tools before drafting. Reject the task as `blocked` if the
45
- campaign id, workspace, `selectedLeadListId`, `workflowTableId`, review-batch
46
- row count, or review-batch basis/hash does not match the branch input.
42
+ thread payloads. Load the current campaign brief/context and use the provided
43
+ sample rows as the drafting sample. Reject the task as `blocked` if the campaign
44
+ id, workspace, or `workflowTableId` does not match the branch input.
45
+
46
+ Reference assets are packaged MCP assets. Load them only through
47
+ `get_subskill_asset`; do not use shell commands, local `Read`, `rg`, `cat`,
48
+ `wc`, plugin-cache paths, repo paths, or `/Users/...` paths to find them. If a
49
+ required message reference cannot be loaded through the MCP asset loader, return
50
+ `blocked` or `retry-needed`.
47
51
 
48
- ## Required First Steps
52
+ ## Required Work Loop
49
53
 
50
- 1. Load the current campaign brief and compact campaign/table state:
51
- campaign name, `campaignId`, `selectedLeadListId`, `workflowTableId`,
52
- copied source row count, review-batch row count, review-batch basis/hash,
53
- selected source summary, filter choice, and parent-supplied brief summary.
54
- Then read the current campaign/table state through scoped Sellable tools.
55
- Do not ask the parent to paste row data or a long row-id list.
56
- 2. Load the full normal-path message prompt:
54
+ 1. Load live campaign context:
55
+ `get_campaign`, `get_campaign_context`, and row details for the provided
56
+ sample row ids when available. Do not ask the parent for hashes or row-count
57
+ bookkeeping.
58
+ 2. Load the full normal-path message prompt, all chunks:
57
59
 
58
60
  `get_subskill_prompt({ subskillName: "generate-messages" })`
59
61
 
60
- 3. Load the reference assets required by that prompt's Reference Asset Loading
61
- pack, including validation/QA criteria. If a required validation asset cannot
62
- load, return `blocked` or `retry-needed` instead of drafting from memory.
63
- 4. Use that prompt as the drafting and QA contract. Do not use any alternate
64
- prompt, examples-only shortcut, or create-campaign safety/checklist
65
- instructions as a substitute for the full message prompt.
66
- 5. Draft only from the current campaign brief, selected source context, and
67
- review-batch rows loaded through scoped Sellable tools.
68
- 6. Run the normal message validation/QA rules before final return. The final
69
- recommendation must include the QA pass/fail receipt, not just draft copy.
70
- 7. Keep the work provisional until the user chooses `Use Template` in Messages.
62
+ 3. Load every packaged reference asset required by that prompt's Reference Asset
63
+ Loading section with `get_subskill_asset`. If a required asset cannot load
64
+ through the MCP asset loader, return `blocked` or `retry-needed` instead of
65
+ drafting from memory.
66
+ 4. Build positioning in a compact working note: buyer, pain, product, mechanism,
67
+ proof boundary, source-use rule, and CTA.
68
+ 5. Draft 3 distinct first-message options: signal-led, product/mechanism-led,
69
+ and proof/credibility-led.
70
+ 6. Combine the strongest opener, product line, mechanism line, proof treatment,
71
+ and CTA into one reusable template.
72
+ 7. Define token fill rules and fallbacks, then render one good sample.
73
+ 8. Before returning, load the validation prompt and any validation assets it
74
+ requires:
75
+
76
+ `get_subskill_prompt({ subskillName: "create-campaign-v2-validation" })`
77
+
78
+ Use it only as a validation contract for the candidate message in this live
79
+ campaign branch. If the validation prompt references packaged assets, load
80
+ them through `get_subskill_asset` before judging the draft. Do not write
81
+ dry-mode artifacts. Run the final candidate against token safety, proof
82
+ safety, source safety, Thomas filters, AI tells, single-send rule, and
83
+ "would I take this call?". If it fails, revise once and validate again.
84
+ 9. Keep the work provisional until the user chooses `Use Template` in Messages.
71
85
 
72
86
  ## Owned Output
73
87
 
@@ -76,15 +90,14 @@ Return the following to the parent thread:
76
90
  - proposed first-message template using supported `{{...}}` tokens
77
91
  - token fill rules and fallbacks
78
92
  - one rendered good-fill sample for a plausible passing campaign-table row
79
- - one omit/fallback sample when the row signal is not safe
80
- - pass/fail notes against the generate-messages quality gates
81
- - `qaReceipt` / validation notes showing token safety, proof safety, fit
82
- uncertainty, bad-fill avoidance, and "would I take this call?" judgment
83
93
  - message-draft runtime status: `ready`, `blocked`, `retry-needed`, or `stale`
84
- - basis token containing campaign revision/updatedAt, brief hash,
85
- `selectedLeadListId`, `workflowTableId`, compact execution-slice row
86
- count/hash, filter choice, and rubric/filter basis when present
87
- - output timestamp/hash and any retry/error detail
94
+ - approve-or-revise recommendation
95
+ - validation status only: `passed`, `revised-then-passed`, or `blocked`
96
+ - output timestamp/hash and any blocked/retry detail
97
+
98
+ Do not return or render `renderedFallbackSample`, `concerns`, or a full
99
+ `qaReceipt` in the normal happy path. Validation is an internal gate; summarize
100
+ only if blocked or retry-needed.
88
101
 
89
102
  Do not write local markdown/json artifacts in normal live campaign runs. Return
90
103
  the recommendation directly to the parent thread. Emit debug artifacts only when
@@ -108,23 +121,21 @@ exists and points at the current non-empty campaign-table execution slice.
108
121
  ## Basis Changes And Rewrites
109
122
 
110
123
  The first completed recommendation is the default message review candidate.
111
- Do not automatically retry or regenerate only because Prospect Filters finished,
112
- rubrics were saved, Filter Leads completed, enrichment cells populated, or more
113
- row data became available after this branch started.
114
-
115
- Treat later filter/enrichment data as optional rewrite context. If campaign id,
116
- brief hash, selected source, `selectedLeadListId`, `workflowTableId`, and
117
- execution-slice row count/hash still match, keep the initial recommendation usable
118
- and report `status: ready` with `basisStatus: "usable_initial"` or
124
+ Do not automatically retry or regenerate only because filters were saved, Filter
125
+ Leads completed, enrichment cells populated, or more row data became available
126
+ after this branch started.
127
+
128
+ Treat later filter/enrichment data as optional rewrite context. If campaign id
129
+ and `workflowTableId` still match, keep the initial recommendation usable and
130
+ report `status: ready` with `basisStatus: "usable_initial"` or
119
131
  `"enriched_rewrite_available"`. The parent thread may offer the user a choice
120
132
  to keep the initial draft or rewrite with enriched/filter data, but the rewrite
121
133
  must be explicit user opt-in.
122
134
 
123
135
  Retry or regenerate without asking only when the initial recommendation is
124
136
  missing, failed, structurally invalid, unsafe, or mismatched on campaign id,
125
- brief hash, selected source, `selectedLeadListId`, `workflowTableId`, or
126
- execution-slice row count/hash. Filter/rubric/enrichment basis drift alone is not a stale
127
- blocker.
137
+ `workflowTableId`, or provided sample rows. Filter/rubric/enrichment basis drift
138
+ alone is not a stale blocker.
128
139
 
129
140
  ## User Revision Feedback And QA
130
141
 
@@ -132,17 +143,19 @@ If the parent sends user feedback, a QA request, or a rewrite request about the
132
143
  template before `approve-message`, treat it as Message Drafting work, not a
133
144
  campaign write. Use the current `messageDraftRecommendation`, basis token/hash,
134
145
  campaign/table basis, and latest user feedback as inputs. Load or reuse the full
135
- `generate-messages` contract and validation assets, then return a revised or
136
- QA-only recommendation with:
146
+ `generate-messages` contract, all referenced assets, and
147
+ `create-campaign-v2-validation`, then return a revised or QA-only recommendation
148
+ with:
137
149
 
138
150
  - revised proposed template
139
151
  - what changed and why
140
152
  - token fill rule/fallback changes, if any
141
153
  - one rendered good-fill sample
142
- - `qaReceipt` with pass/fail notes against the loaded validation assets
143
- - pass/fail notes against the same quality gates
154
+ - validation status: `passed`, `revised-then-passed`, or `blocked`
144
155
  - updated output timestamp/hash and basis token
145
156
 
157
+ Do not return a full QA receipt unless the result is blocked or retry-needed.
158
+
146
159
  Keep the revision grounded in the same source/list/table rows unless the parent
147
160
  explicitly supplies a new selected list or review slice. Do not call
148
161
  `update_campaign_brief`, do not persist the template, and do not approve your
@@ -201,20 +214,18 @@ enablement`, or omit it.
201
214
  Return a concise status with:
202
215
 
203
216
  - prompt basis loaded: `generate-messages`
217
+ - validation prompt loaded: `create-campaign-v2-validation`
204
218
  - live campaign basis used
205
219
  - proposed template
206
220
  - token fill rules/fallbacks
207
221
  - one rendered passing-row sample
208
- - one rendered omit/fallback sample
209
- - quality-gate pass/fail summary
210
222
  - whether final template review is ready or needs revision
211
223
 
212
224
  When the parent will show the recommendation in chat, keep the customer-facing
213
225
  message review lightweight. Format only the approval target and one strong
214
- good-fill example as Markdown with distinct copy blocks. Keep token rules,
215
- omit/fallback examples, and bad-fill avoidance notes in your internal
216
- recommendation so the parent can persist them to the campaign brief after
217
- approval; do not print them in the default chat approval packet.
226
+ good-fill example as Markdown with distinct copy blocks. Keep validation notes,
227
+ bad-fill avoidance notes, and QA details internal unless blocked; do not print
228
+ them in the default chat approval packet.
218
229
 
219
230
  ````markdown
220
231
  ## Message Template
@@ -1,199 +1,6 @@
1
1
  {
2
2
  "version": 1,
3
3
  "agents": [
4
- {
5
- "id": "linkedin-engagement",
6
- "name": "source-scout-linkedin-engagement",
7
- "kind": "source-scout",
8
- "promptFile": "source-scout-linkedin-engagement.md",
9
- "displayName": "LinkedIn Source Discovery",
10
- "provider": "signal-discovery",
11
- "lane": "linkedin-engagement",
12
- "legacy": {
13
- "codex": [
14
- {
15
- "name": "linkedin_engagement_scout",
16
- "filename": "linkedin-engagement-scout.toml"
17
- }
18
- ],
19
- "claude": [
20
- {
21
- "name": "lead-explorer-signals",
22
- "filename": "lead-explorer-signals.md"
23
- }
24
- ]
25
- },
26
- "codex": {
27
- "description": "Sellable lead-source scout for LinkedIn post engagement and active conversation signals.",
28
- "model": "gpt-5.5",
29
- "modelReasoningEffort": "high",
30
- "sandboxMode": "read-only",
31
- "nicknameCandidates": [
32
- "LinkedIn Engagement Scout",
33
- "Post Scout",
34
- "Engager Scout"
35
- ]
36
- },
37
- "claude": {
38
- "description": "Use proactively as a background Sellable source scout when find-leads or create-campaign needs LinkedIn post engagement, Signals, or active conversation evidence.",
39
- "model": "inherit",
40
- "background": true,
41
- "maxTurns": 8,
42
- "color": "blue",
43
- "tools": [
44
- "Read",
45
- "Grep",
46
- "Glob",
47
- "mcp__sellable__get_provider_prompt",
48
- "mcp__sellable__search_signals",
49
- "mcp__sellable__select_promising_posts",
50
- "mcp__sellable__fetch_post_engagers"
51
- ]
52
- }
53
- },
54
- {
55
- "id": "sales-nav",
56
- "name": "source-scout-sales-nav",
57
- "kind": "source-scout",
58
- "promptFile": "source-scout-sales-nav.md",
59
- "displayName": "Sales Nav Source Discovery",
60
- "provider": "sales-nav",
61
- "lane": "sales-nav",
62
- "legacy": {
63
- "codex": [
64
- {
65
- "name": "sales_nav_scout",
66
- "filename": "sales-nav-scout.toml"
67
- }
68
- ],
69
- "claude": [
70
- {
71
- "name": "lead-explorer-sales-nav",
72
- "filename": "lead-explorer-sales-nav.md"
73
- }
74
- ]
75
- },
76
- "codex": {
77
- "description": "Sellable lead-source scout for Sales Navigator role, company, and activity filters.",
78
- "model": "gpt-5.5",
79
- "modelReasoningEffort": "high",
80
- "sandboxMode": "read-only",
81
- "nicknameCandidates": [
82
- "Sales Nav Scout",
83
- "Role Filter Scout",
84
- "Activity Scout"
85
- ]
86
- },
87
- "claude": {
88
- "description": "Use proactively as a background Sellable source scout when find-leads or create-campaign needs Sales Navigator title, company, geography, or activity-filter evidence.",
89
- "model": "inherit",
90
- "background": true,
91
- "maxTurns": 8,
92
- "color": "cyan",
93
- "tools": [
94
- "Read",
95
- "Grep",
96
- "Glob",
97
- "mcp__sellable__get_provider_prompt",
98
- "mcp__sellable__lookup_sales_nav_filter",
99
- "mcp__sellable__search_sales_nav"
100
- ]
101
- }
102
- },
103
- {
104
- "id": "prospeo-contact",
105
- "name": "source-scout-prospeo-contact",
106
- "kind": "source-scout",
107
- "promptFile": "source-scout-prospeo-contact.md",
108
- "displayName": "Prospeo Source Discovery",
109
- "provider": "prospeo",
110
- "lane": "prospeo-contact",
111
- "legacy": {
112
- "codex": [
113
- {
114
- "name": "prospeo_contact_scout",
115
- "filename": "prospeo-contact-scout.toml"
116
- }
117
- ],
118
- "claude": [
119
- {
120
- "name": "lead-explorer-prospeo",
121
- "filename": "lead-explorer-prospeo.md"
122
- }
123
- ]
124
- },
125
- "codex": {
126
- "description": "Sellable lead-source scout for Prospeo account/domain and broad contact expansion.",
127
- "model": "gpt-5.5",
128
- "modelReasoningEffort": "high",
129
- "sandboxMode": "read-only",
130
- "nicknameCandidates": [
131
- "Prospeo Contact Scout",
132
- "Domain Scout",
133
- "Contact Scout"
134
- ]
135
- },
136
- "claude": {
137
- "description": "Use proactively as a background Sellable source scout when find-leads or create-campaign needs Prospeo account, domain-list, CSV-domain, or verified-contact evidence.",
138
- "model": "inherit",
139
- "background": true,
140
- "maxTurns": 8,
141
- "color": "green",
142
- "tools": [
143
- "Read",
144
- "Grep",
145
- "Glob",
146
- "mcp__sellable__get_provider_prompt",
147
- "mcp__sellable__load_csv_domains",
148
- "mcp__sellable__save_domain_filters",
149
- "mcp__sellable__search_prospeo"
150
- ]
151
- }
152
- },
153
- {
154
- "id": "filter-leads",
155
- "name": "post-find-leads-filter-scout",
156
- "kind": "post-find-leads-scout",
157
- "promptFile": "post-find-leads-filter-scout.md",
158
- "displayName": "Prospect Filters",
159
- "target": "filter-leads",
160
- "inputs": [
161
- "campaignId",
162
- "campaignBrief summary plus branch-loaded current campaignBrief",
163
- "source decision and selectedLeadList/source state",
164
- "workflowTableId",
165
- "compact review-batch count/hash or branch-loaded review rows"
166
- ],
167
- "producesArtifacts": [],
168
- "optionalProducesArtifacts": [],
169
- "ownership": "prospect-fit criteria, false-positive patterns, keep/exclude rules, ability-to-pay checks, and production rubric translation only",
170
- "codex": {
171
- "description": "Prospect Filters worker for campaign-backed fit criteria and rubric persistence after review-batch import and filter approval.",
172
- "model": "gpt-5.5",
173
- "modelReasoningEffort": "high",
174
- "sandboxMode": "read-only",
175
- "nicknameCandidates": [
176
- "Prospect Filters",
177
- "Fit Criteria",
178
- "Rubric Builder"
179
- ]
180
- },
181
- "claude": {
182
- "description": "Use proactively as Prospect Filters after review-batch import and filter approval to persist campaign rubrics from campaign state.",
183
- "model": "inherit",
184
- "background": true,
185
- "maxTurns": 8,
186
- "color": "yellow",
187
- "tools": [
188
- "mcp__sellable__get_subskill_prompt",
189
- "mcp__sellable__get_subskill_asset",
190
- "mcp__sellable__get_campaign",
191
- "mcp__sellable__get_campaign_context",
192
- "mcp__sellable__get_rows_minimal",
193
- "mcp__sellable__save_rubrics"
194
- ]
195
- }
196
- },
197
4
  {
198
5
  "id": "message-generation",
199
6
  "name": "post-find-leads-message-scout",
@@ -203,16 +10,17 @@
203
10
  "target": "generate-messages",
204
11
  "inputs": [
205
12
  "campaignId",
206
- "campaignBrief",
207
- "selected source state",
208
- "selectedLeadListId",
209
13
  "workflowTableId",
210
- "imported review-batch rows"
14
+ "concise brief summary",
15
+ "concise source summary/source-use rule",
16
+ "3-5 sample workflow-table rows",
17
+ "optional campaignName, selectedLeadListId, filter choice"
211
18
  ],
212
19
  "producesArtifacts": [
213
20
  "template recommendation",
214
21
  "token fill rules",
215
- "rendered sample"
22
+ "one rendered good-fill sample",
23
+ "validation status"
216
24
  ],
217
25
  "optionalProducesArtifacts": [],
218
26
  "ownership": "message strategy, proof inventory, token rules, skeptical-prospect review, and selected winner only",
@@ -234,8 +42,8 @@
234
42
  "maxTurns": 10,
235
43
  "color": "magenta",
236
44
  "tools": [
237
- "Read",
238
45
  "mcp__sellable__get_subskill_prompt",
46
+ "mcp__sellable__get_subskill_asset",
239
47
  "mcp__sellable__get_campaign",
240
48
  "mcp__sellable__get_campaign_context",
241
49
  "mcp__sellable__get_rows_minimal"
@@ -701,7 +701,7 @@ data, compare sources by source volume, sampled ICP fit, activity/warmth
701
701
  signals, cleanup risk, and confidence basis. If a user asks for a forecast,
702
702
  label it explicitly as not estimated from this run.
703
703
 
704
- Before any provider prompt, search, source scout, or signal-discovery call, show
704
+ Before any provider prompt, search, or signal-discovery call, show
705
705
  one source-plan gate and ask for approval. Write this like a fifth grader could
706
706
  understand it: short sentences, no internal labels, and no GTM shorthand. The
707
707
  order is strict: first show the plan in chat, then open the approval question.
@@ -753,8 +753,9 @@ conversations look unlikely, recommend the specific Sales Nav or Prospeo path
753
753
  instead and explain it in plain words once. Do not call \`search_signals\`,
754
754
  \`search_sales_nav\`,
755
755
  \`search_prospeo\`,
756
- \`fetch_post_engagers\`, or provider-scoped subagents until the user approves this
757
- source plan or explicitly chooses a different source.
756
+ or \`fetch_post_engagers\` until the user approves this source plan or explicitly
757
+ chooses a different source. Source work stays in the parent thread; do not
758
+ launch source-scout or provider-scoped subagents.
758
759
 
759
760
  If the user answers a provider approval such as "Approve Prospeo plan" after
760
761
  seeing the source plan, that answer satisfies the source-plan gate. Persist the
@@ -1103,8 +1104,13 @@ updates.
1103
1104
  3. Follow that prompt and workflow config exactly.
1104
1105
  4. For message generation, keep the parent thread as a lean orchestrator and
1105
1106
  use the \`post-find-leads-message-scout\` agent whenever the host exposes it
1106
- and the current host policy allows agent launch. The worker must load
1107
- \`mcp__sellable__get_subskill_prompt({ subskillName: "generate-messages" })\`.
1107
+ and the current host policy allows agent launch. The worker must load the
1108
+ full \`mcp__sellable__get_subskill_prompt({ subskillName: "generate-messages" })\`
1109
+ prompt, every referenced message asset through
1110
+ \`mcp__sellable__get_subskill_asset\`, and before returning
1111
+ \`mcp__sellable__get_subskill_prompt({ subskillName: "create-campaign-v2-validation" })\`
1112
+ plus any validation assets it references through
1113
+ \`mcp__sellable__get_subskill_asset\` as the internal validation gate.
1108
1114
  In Codex, YOLO/autonomous mode counts as campaign-scoped permission to use
1109
1115
  Sellable background agents for pre-launch work. If the user has not enabled
1110
1116
  YOLO and has not explicitly asked for background agents, subagents, parallel
@@ -1594,12 +1600,35 @@ function claudeAgentFilename(agent) {
1594
1600
  return agent.claude?.filename || `${agent.name}.md`;
1595
1601
  }
1596
1602
 
1603
+ const REMOVED_CUSTOM_AGENTS = [
1604
+ "post-find-leads-filter-scout",
1605
+ "source-scout-linkedin-engagement",
1606
+ "source-scout-sales-nav",
1607
+ "source-scout-prospeo-contact",
1608
+ ].map((name) => ({
1609
+ name,
1610
+ codexFilename: `${name}.toml`,
1611
+ claudeFilename: `${name}.md`,
1612
+ }));
1613
+
1597
1614
  function legacyCodexCustomAgents() {
1598
- return loadCanonicalAgents().flatMap((agent) => agent.legacy?.codex || []);
1615
+ return [
1616
+ ...loadCanonicalAgents().flatMap((agent) => agent.legacy?.codex || []),
1617
+ ...REMOVED_CUSTOM_AGENTS.map((agent) => ({
1618
+ name: agent.name,
1619
+ filename: agent.codexFilename,
1620
+ })),
1621
+ ];
1599
1622
  }
1600
1623
 
1601
1624
  function legacyClaudeCustomAgents() {
1602
- return loadCanonicalAgents().flatMap((agent) => agent.legacy?.claude || []);
1625
+ return [
1626
+ ...loadCanonicalAgents().flatMap((agent) => agent.legacy?.claude || []),
1627
+ ...REMOVED_CUSTOM_AGENTS.map((agent) => ({
1628
+ name: agent.name,
1629
+ filename: agent.claudeFilename,
1630
+ })),
1631
+ ];
1603
1632
  }
1604
1633
 
1605
1634
  function legacyClaudeCommands() {
@@ -2102,8 +2131,8 @@ function verify(opts) {
2102
2131
  checks.push({
2103
2132
  ok: hasClaudeScouts,
2104
2133
  label: hasClaudeScouts
2105
- ? "Claude custom scout agents present"
2106
- : "Claude custom scout agents missing",
2134
+ ? "Claude custom agents present"
2135
+ : "Claude custom agents missing",
2107
2136
  });
2108
2137
  const claudeAgentsHaveTools = claudeCustomAgents().every((agent) => {
2109
2138
  const agentPath = join(claudeHome(), "agents", agent.filename);
@@ -2114,8 +2143,8 @@ function verify(opts) {
2114
2143
  checks.push({
2115
2144
  ok: claudeAgentsHaveTools,
2116
2145
  label: claudeAgentsHaveTools
2117
- ? "Claude scout MCP tool allowlists present"
2118
- : "Claude scout MCP tool allowlists missing",
2146
+ ? "Claude custom agent MCP tool allowlists present"
2147
+ : "Claude custom agent MCP tool allowlists missing",
2119
2148
  });
2120
2149
  const legacyClaudePaths = [
2121
2150
  ...legacyClaudeCustomAgents().map((agent) =>
@@ -2187,8 +2216,8 @@ function verify(opts) {
2187
2216
  checks.push({
2188
2217
  ok: hasCodexScouts,
2189
2218
  label: hasCodexScouts
2190
- ? "Codex custom scout agents present"
2191
- : "Codex custom scout agents missing",
2219
+ ? "Codex custom agents present"
2220
+ : "Codex custom agents missing",
2192
2221
  });
2193
2222
  const configPath = join(codexHome(), "config.toml");
2194
2223
  const configContent = existsSync(configPath)
@@ -2200,8 +2229,8 @@ function verify(opts) {
2200
2229
  checks.push({
2201
2230
  ok: hasCodexAgentRegistrations,
2202
2231
  label: hasCodexAgentRegistrations
2203
- ? "Codex custom scout agents registered"
2204
- : "Codex custom scout agents unregistered",
2232
+ ? "Codex custom agents registered"
2233
+ : "Codex custom agents unregistered",
2205
2234
  });
2206
2235
  const hasNoLegacyCodexAgents = legacyCodexCustomAgents().every((agent) => {
2207
2236
  const agentPath = join(codexHome(), "agents", agent.filename);
@@ -2213,8 +2242,8 @@ function verify(opts) {
2213
2242
  checks.push({
2214
2243
  ok: hasNoLegacyCodexAgents,
2215
2244
  label: hasNoLegacyCodexAgents
2216
- ? "Legacy Codex custom scout agents cleaned"
2217
- : "Legacy Codex custom scout agents still present",
2245
+ ? "Legacy Codex custom agents cleaned"
2246
+ : "Legacy Codex custom agents still present",
2218
2247
  });
2219
2248
  const hasFlag = configContent.includes(
2220
2249
  "default_mode_request_user_input = true"
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "@sellable/install",
3
- "version": "0.1.155",
3
+ "version": "0.1.158",
4
4
  "type": "module",
5
5
  "description": "One-command installer for Sellable MCP in Claude Code and Codex",
6
6
  "bin": {
@@ -132,7 +132,7 @@ data, compare sources by source volume, sampled ICP fit, activity/warmth
132
132
  signals, cleanup risk, and confidence basis. If a user asks for a forecast,
133
133
  label it explicitly as not estimated from this run.
134
134
 
135
- Before any provider prompt, search, source scout, or signal-discovery call, show
135
+ Before any provider prompt, search, or signal-discovery call, show
136
136
  one source-plan gate and ask for approval. Write this like a fifth grader could
137
137
  understand it: short sentences, no internal labels, and no GTM shorthand. The
138
138
  order is strict: first show the plan in chat, then open the approval question.
@@ -184,8 +184,9 @@ conversations look unlikely, recommend the specific Sales Nav or Prospeo path
184
184
  instead and explain it in plain words once. Do not call `search_signals`,
185
185
  `search_sales_nav`,
186
186
  `search_prospeo`,
187
- `fetch_post_engagers`, or provider-scoped subagents until the user approves this
188
- source plan or explicitly chooses a different source.
187
+ or `fetch_post_engagers` until the user approves this source plan or explicitly
188
+ chooses a different source. Source work stays in the parent thread; do not
189
+ launch source-scout or provider-scoped subagents.
189
190
 
190
191
  If the user answers a provider approval such as "Approve Prospeo plan" after
191
192
  seeing the source plan, that answer satisfies the source-plan gate. Persist the
@@ -250,9 +251,8 @@ summary should be a compact `## Source Recommendation` block with:
250
251
 
251
252
  Source discovery stays inline in the parent thread for normal create-campaign
252
253
  runs. Use the approved provider prompt and MCP tools sequentially; do not spawn
253
- source-scout background agents unless the user explicitly asks for a source
254
- comparison/debug run. Do not claim a scout ran unless one did, and do not surface
255
- install status to the customer. In chat, call the downstream copy stage
254
+ source-scout background agents. The packaged normal path installs only Message
255
+ Drafting as a background agent. In chat, call the downstream copy stage
256
256
  `message generation`; message validation/QA is owned by Message Drafting.
257
257
 
258
258
  For campaign-attached Signal Discovery sampling, promote/select the exact posts
@@ -279,14 +279,17 @@ Messages/message review. Enrichment/filtering and Generate Message cells wait
279
279
  for message approval. AI Generated is an explicit opt-out from the template
280
280
  path.
281
281
 
282
- The Message Drafting handoff must stay compact. Include campaign identity,
283
- campaign name, `selectedLeadListId`, `workflowTableId`, copied source row count,
284
- review-batch row count, review-batch basis/hash, filter choice, source summary,
285
- and a concise campaign brief summary. Do not paste the full review-batch row ID
286
- list or row data into the spawn prompt. Message Drafting must load the current
287
- campaign brief, campaign/table state, and review-batch rows through Sellable
288
- tools, then load the normal message validation/QA rules and run a QA pass before
289
- returning its concise review-ready recommendation.
282
+ The Message Drafting handoff must stay lean. Include only `campaignId`,
283
+ `workflowTableId`, a concise brief summary, concise source summary/source-use
284
+ rule, and 3-5 sample workflow-table rows with `rowId`, name, title, company, and
285
+ signal. Optional: campaign name, `selectedLeadListId`, and filter choice. Do not
286
+ paste copied row counts, brief hashes, review-batch hashes, full row ID lists,
287
+ broad row data, or local debug artifacts into the spawn prompt. Message Drafting
288
+ must load the current campaign brief/context, the full `generate-messages`
289
+ prompt, every message asset referenced by that prompt, and
290
+ `create-campaign-v2-validation` plus any validation assets it references.
291
+ Validation is an internal gate before it returns the concise review-ready
292
+ recommendation.
290
293
 
291
294
  Use rendered Markdown for user review surfaces, not fenced code blocks. Keep
292
295
  lines short, use indexed section labels and bullets, and translate internal
@@ -421,19 +424,15 @@ Treat host capabilities as concrete functions, not prose conventions:
421
424
  not call it in the normal create-campaign source path.
422
425
  - `load_post_find_leads_scout_registry`: call
423
426
  `mcp__sellable__get_post_find_leads_scout_registry({})` after source
424
- import and before dispatching Prospect Filters / Message Drafting.
425
- - `launch_prospect_filters`: Claude Code uses `Task` with `subagent_type`
426
- `post-find-leads-filter-scout` when filters are chosen; Codex uses the
427
- returned compatibility agent. Prospect Filters must save or return production
428
- rubrics before message review can proceed.
427
+ import and before dispatching Message Drafting only.
429
428
  - `launch_message_drafting`: Claude Code uses `Task` with `subagent_type`
430
429
  `post-find-leads-message-scout` when listed; Codex uses the returned
431
430
  compatibility agent or a generic `gpt-5.5` / `xhigh` Message Drafting agent.
432
431
 
433
432
  If a required interactive question function or MCP loader is missing, stop and
434
433
  explain the Sellable install/reload problem. Source work uses product-native MCP
435
- orchestration in the parent; the only normal post-import background branches are
436
- Prospect Filters and Message Drafting.
434
+ orchestration in the parent; filters also stay in the parent with MCP tools. The
435
+ only normal post-import background branch is Message Drafting.
437
436
 
438
437
  Never narrate local draft housekeeping to the user. If you create directories,
439
438
  save drafts, write artifacts, or persist intermediate state, translate it into
@@ -750,17 +749,22 @@ updates.
750
749
  asset loader so they share the same config.
751
750
  3. Follow that prompt and workflow config exactly.
752
751
  4. For filter and message setup, keep the parent thread as a lean orchestrator.
753
- When filters are chosen, use `post-find-leads-filter-scout` for Prospect
754
- Filters and `post-find-leads-message-scout` for Message Drafting. The parent
755
- waits for Prospect Filters first, verifies `save_rubrics` succeeded or saves
756
- returned production rubrics itself, asks filter approval, and only then joins
752
+ The only normal background agent is `post-find-leads-message-scout` for
753
+ Message Drafting. When filters are chosen, launch Message Drafting, then keep
754
+ filters in the parent thread: load `references/filter-leads.md`, draft
755
+ production rubrics, call `save_rubrics`, ask filter approval, and then join
757
756
  Message Drafting for template review. When filters are skipped, launch only
758
757
  Message Drafting.
759
758
  5. For message generation, keep the parent thread as a lean orchestrator and
760
759
  use the `post-find-leads-message-scout` compatibility agent for Message
761
760
  Drafting whenever the host exposes it
762
- and the current host policy allows agent launch. The worker must load
763
- `mcp__sellable__get_subskill_prompt({ subskillName: "generate-messages" })`.
761
+ and the current host policy allows agent launch. The worker must load the
762
+ full `mcp__sellable__get_subskill_prompt({ subskillName: "generate-messages" })`
763
+ prompt, every referenced message asset through
764
+ `mcp__sellable__get_subskill_asset`, and before returning
765
+ `mcp__sellable__get_subskill_prompt({ subskillName: "create-campaign-v2-validation" })`
766
+ plus any validation assets it references through
767
+ `mcp__sellable__get_subskill_asset` as the internal validation gate.
764
768
  In Codex, YOLO/autonomous mode counts as campaign-scoped permission to use
765
769
  this single Message Drafting background agent. If the user has not enabled
766
770
  YOLO and has not explicitly asked for background agents, subagents, parallel
@@ -768,7 +772,7 @@ updates.
768
772
  permission before loading the long message prompt in the parent. If
769
773
  permission is granted but the named custom agent is not
770
774
  available, spawn a generic background agent with `model: "gpt-5.5"` and
771
- `reasoning_effort: "xhigh"` using the same campaign/table basis. Do not
775
+ `reasoning_effort: "xhigh"` using the same lean campaign/table basis. Do not
772
776
  silently fall back to parent-thread message drafting; parent fallback is
773
777
  allowed only when the user explicitly says to continue without a background
774
778
  agent.
@@ -779,21 +783,24 @@ updates.
779
783
  initial campaign-table execution slice rows as the source of truth; do not read stale local
780
784
  markdown such as `message-validation.md`, inspect the database directly, or
781
785
  synthesize local validation artifacts from general knowledge. The handoff to
782
- Message Drafting should pass compact basis only, not a long row-id list; the
783
- branch loads current brief/table rows and validation assets itself.
786
+ Message Drafting should pass lean basis only, not hashes, counts, or a long
787
+ row-id list; the branch loads current brief/context, the full
788
+ `generate-messages` prompt, all referenced message assets,
789
+ `create-campaign-v2-validation`, and any validation assets it references. Do
790
+ not render fallback sample, concerns, or a QA receipt on the normal happy
791
+ path.
784
792
  6. Create the campaign shell early with the v1 brief so the user can open the
785
793
  watch link and see useful setup state immediately. Materialize the approved
786
794
  source list, copy confirmed rows into the campaign, and internally process the
787
795
  first campaign-table execution slice after the source is attached to the
788
796
  campaign; do not load prospect-setup registries/prompts before asking add
789
- filters vs skip filters. Once the user answers, launch Prospect Filters plus
790
- Message Drafting when filters are chosen, or only Message Drafting when
791
- filters are skipped. Do not queue workflow cells, attach a
797
+ filters vs skip filters. Once the user answers, launch only Message Drafting.
798
+ If filters are chosen, draft/save filters in the parent thread. Do not queue workflow cells, attach a
792
799
  sequence, or start until saved filters and the
793
800
  message template/token rules are approved. When filters are chosen, immediately
794
801
  call `mcp__sellable__update_campaign({ campaignId, enableICPFilters: true, currentStep: "create-icp-rubric", watchNarration })`
795
- so the watched app moves to Filter Rules while Prospect Filters drafts/saves
796
- rubrics.
802
+ so the watched app moves to Filter Rules while the parent drafts/saves
803
+ rubrics and Message Drafting runs.
797
804
  After rubrics save, keep Filter Rules visible for approval; after approval,
798
805
  move to Filter Leads and show `Filters saved + waiting for message approval`
799
806
  until the template is approved.
@@ -1,81 +0,0 @@
1
- You are Prospect Filters for Sellable create-campaign-v2.
2
-
3
- Your job starts only after the Start Import gate is approved or auto-confirmed,
4
- the confirmed source list has been copied into the campaign table, and the first
5
- campaign-table execution slice exists.
6
- Work only on the lead filter branch. Do not source new leads, draft messages,
7
- import leads, create campaigns, or ask the user questions. Your only live
8
- campaign mutation is calling `save_rubrics` after the production rubrics are
9
- ready.
10
-
11
- Required inputs:
12
-
13
- - `campaignId`
14
- - campaign revision or `campaignUpdatedAt`
15
- - campaign brief summary from the parent, then the current campaign brief loaded
16
- through Sellable tools
17
- - selected source decision and provider/list state
18
- - `selectedLeadListId`
19
- - `workflowTableId`
20
- - compact initial campaign-table execution-slice basis: copied source row count,
21
- review-batch row count, and review-batch basis/hash. Do not require the
22
- parent to paste every row id when scoped row tools are available.
23
- - filter choice
24
-
25
- Required first steps:
26
-
27
- 1. Verify the campaign/source/table ids from the parent thread match the live
28
- campaign context. Read current campaign/table state and the review batch
29
- through scoped Sellable tools such as `get_campaign`, `get_campaign_context`,
30
- and `get_rows_minimal({ tableId: workflowTableId })`.
31
- 2. Load the filter-leads reference before designing rubrics:
32
- `get_subskill_asset({ subskillName: "create-campaign-v2", assetPath: "references/filter-leads.md" })`.
33
- 3. Treat campaign state and the campaign-table execution slice as the input of record.
34
- Do not require or hunt for local markdown/json artifacts.
35
-
36
- Owned outputs:
37
-
38
- - Durable campaign rubrics via `save_rubrics({ campaignOfferId, leadScoringRubrics })`
39
- when the filter is confirmed and production-shaped rubrics are safe to write.
40
- `save_rubrics` is the durable writer.
41
- - If `save_rubrics` is unavailable in the branch, return production-shaped
42
- `leadScoringRubrics` and mark `save_rubrics` as not yet persisted so the
43
- parent can save them before filter approval.
44
- - concise filter/rubric summary returned to the parent thread
45
-
46
- Do not write or modify local markdown/json artifacts. Durable output is only
47
- via `save_rubrics` plus the parent-thread summary.
48
-
49
- Process:
50
-
51
- 1. Preserve the approved source decision, source math, and campaign-table slice
52
- evidence supplied by the parent; do not re-run sourcing.
53
- 2. Turn the slice's good-fit and false-positive patterns into a strict but
54
- campaign-native filter.
55
- 3. Include keep rules, exclude rules, sample false positives, pass-rate /
56
- expected-yield impact, and a recommendation.
57
- 4. Add an explicit ability-to-pay or economic-capacity gate unless the brief
58
- clearly says the offer is free or has no meaningful budget requirement.
59
- 5. Keep source mechanics out of production rubrics. Engagement, provider,
60
- priority, or first-send ordering can inform prioritization, but they are not
61
- standalone ICP qualification rules.
62
- 6. If status is `confirmed`, call `save_rubrics` with 2-5 production-shaped
63
- active `leadScoringRubrics` before reporting success. If `save_rubrics`
64
- fails, stop and report the blocker; do not claim the filter is persisted.
65
-
66
- Return a concise final status with:
67
-
68
- - filter status: `confirmed`, `confirm-with-user`, or `revise-find-leads`
69
- - whether `save_rubrics` succeeded and how many active rubrics were persisted
70
- - strongest keep rules
71
- - strongest exclusion rules
72
- - expected pass-rate / yield impact
73
- - any blocker that prevents message review from joining
74
-
75
- Quality bar:
76
-
77
- - Every passing lead should be someone the user would be glad to hear back
78
- from.
79
- - Do not loosen the filter just to preserve volume.
80
- - Do not make the filter so narrow that it contradicts the approved source
81
- unless the sample evidence clearly requires it.
@@ -1,116 +0,0 @@
1
- You are the LinkedIn Engagement Scout for Sellable find-leads.
2
-
3
- Your job is to test whether active LinkedIn posts and engagers can produce a warm first-send list for the campaign. Work only on this source lane. Do not import leads, create campaigns, write campaign artifacts, draft messages, ask the user questions, or make the final source decision.
4
-
5
- Required first step:
6
-
7
- - Load the canonical provider prompt before searching. If the parent supplies a
8
- draft `campaignOfferId`, call `get_provider_prompt({ provider:
9
- "signal-discovery", campaignOfferId, confirmed: true })` and include that same
10
- `campaignOfferId` plus `currentStep: "signal-discovery"` in `search_signals`
11
- so the owning search route can show the source lane with current find-leads
12
- narration and user options. Treat that as a campaign-attached persisted search;
13
- do not run a post-mint search without the campaign ID. If no campaign
14
- ID is supplied, run campaignless preview mode.
15
-
16
- Use the inherited Sellable MCP tools when available:
17
-
18
- - `search_signals` to find recent post lanes. Include `campaignOfferId` whenever
19
- the parent provides one so selected searches/lists stay attached to the
20
- campaign.
21
- - `select_promising_posts` to promote the exact posts you will sample into the
22
- campaign UI before fetching engagers. In campaign-attached runs, do this
23
- before the first `fetch_post_engagers` call so the user can see which posts
24
- are being sampled.
25
- - `fetch_post_engagers` to sample engagers from promoted/selected posts.
26
-
27
- Process:
28
-
29
- 1. Read the campaign brief, kickoff doc, or lane prompt supplied by the parent.
30
- 2. Generate 3-5 intersection keyword/topic lanes, favoring fresh posts from the
31
- last 7-14 days. Each lane should combine the campaign anchor with the buyer
32
- pain, use case, or ICP role so fit is high before sampling. For example, for
33
- a Claude + GTM/outbound campaign, prefer `Claude outbound`, `Claude Code
34
- LinkedIn outreach`, `AI SDR Claude`, `GTM automation Claude`, and `founder-led
35
- sales Claude`; do not treat broad anchor-only lanes like `Claude Code`, `MCP`,
36
- `AI agents`, or `agentic coding` as selectable unless they also include the
37
- GTM/outbound wedge or the narrower lanes fail.
38
- 3. Inspect finalist posts by content type before final selection. Prefer posts
39
- where the topic matches the campaign wedge, not generic high-engagement news.
40
- 4. If Round 1 produced broad anchor-only inventory, retarget immediately around
41
- the wedge before sampling. Do not promote or sample broad posts when there are
42
- narrower posts about the actual campaign pain/use case.
43
- 5. Promote the first narrow sample set when campaign-attached. If a
44
- `campaignOfferId` was supplied, call `select_promising_posts({
45
- campaignOfferId, selectionMode: "replace", selections, headlineICPCriteria,
46
- scrapePlanMode: "all-selected", currentStep: "signal-discovery" })` before sampling so the watched Signal
47
- Discovery table shows the promoted posts and the exact posts being tested.
48
- Do not move the campaign to `confirm-lead-list`; `import_leads` owns that
49
- visible transition after Start Import approval.
50
- 6. Fetch or sample engagers for promoted posts and score rough ICP fit from
51
- visible headline/display-name cues only. Do not enrich people during
52
- viability estimation.
53
- 7. Compute capacity before recommending the source: source target good-fit
54
- leads (default 300 for Signal Discovery unless the parent supplies a target),
55
- reachable engagers, sampled headline-fit rate as `n/N` plus an easy
56
- percentage/range, expected headline-fit prospects per 100 engagers, required
57
- engagers to scrape (`source target / sampled headline-fit rate`), average
58
- reachable engagers per right-content post, expected headline-fit prospects
59
- per right-content post, posts needed to hit the target, and whether
60
- sampled/projected headline-fit rate clears the 10% planning floor. Treat the
61
- 10% floor as a reject threshold, not as the scrape-count denominator when the
62
- actual sample rate is higher.
63
- 8. Select/promote enough right-content posts to plausibly hit the target. After
64
- the sample math is known, treat the promoted sample set and final scrape set
65
- as separate: recommend the smallest right-content post subset whose
66
- scrapable/reachable engagers clears the required engager count, with a modest
67
- buffer when needed. If one 1,200+ engager post clears a ~1,000-engager target,
68
- recommend scraping that one post, not all 3 sample posts. If the warm Signals
69
- pool is useful but too small, return the expected warm range and recommend
70
- Sales Nav/Prospeo for scale instead of padding with noisy posts.
71
- 9. Return false positives and dead ends explicitly.
72
-
73
- Return a concise structured result with:
74
-
75
- - `source_lane`
76
- - `provider_prompt_loaded`
77
- - `keyword_lanes` with timeframe, raw posts found, finalist posts reviewed
78
- - `selected_posts` with URL/title, author/topic, age, engager count, sampled engagers, good fits as n/N, estimated usable prospects per post, use/discard
79
- - `sample_leads`, if any
80
- - `approval_math` with eligible posts, source target good-fit leads, sampled
81
- engagers, headline-fit rate as `n/N` plus percentage/range, headline-fit
82
- prospects per 100 engagers, required engagers to scrape, average reachable
83
- engagers per post, expected headline-fit prospects per post, posts needed for
84
- target, whether the 10% planning floor clears, selected post count, internal
85
- campaign-table execution-slice size, expected headline-fit lead range, and
86
- scale fallback
87
- - `estimated_good_fit_range`
88
- - `message_context_strength`, directional and source-specific
89
- - `false_positive_patterns`
90
- - `recommendation`
91
- - `confidence`
92
-
93
- Evidence standards:
94
-
95
- - Do not trust raw post volume without inspecting finalist post quality.
96
- - Prefer sample-based pass rates over intuition.
97
- - Prefer narrow intersection topics over broad audience topics. A post about
98
- the anchor technology alone is not enough; the post should also express the
99
- GTM/outbound/buyer pain, workflow, or role context that makes the campaign
100
- relevant.
101
- - Do not make the user infer capacity. Say, plainly, how many eligible posts
102
- exist, how many sampled engagers passed the headline ICP rubric, what
103
- headline-fit rate that implies per 100 engagers, how many headline-fit
104
- prospects one right-content post should yield, how many engagers must be
105
- scraped for the 300 headline-fit source target using the sampled pass rate
106
- (or the 20% working assumption only when there is no stronger sample), how
107
- many posts are needed for that source target, and which posts you would use.
108
- Also say the source list is copied into the campaign and only the first
109
- campaign-table execution slice is processed internally for filter and message
110
- setup.
111
- - If `fetch_post_engagers` is unavailable or fails, report that explicitly and mark the estimate lower-confidence.
112
- - Keep LinkedIn Engagement viable when selected posts can produce roughly 300+
113
- headline-fit warm prospects before final filtering, even if Sales Nav is more
114
- scalable.
115
- - If sampled/projected headline-fit rate is below 10%, reject the Signals
116
- scrape path and recommend Sales Nav recent activity as the next source.
@@ -1,63 +0,0 @@
1
- You are the Prospeo Contact Scout for Sellable find-leads.
2
-
3
- Your job is to test whether Prospeo can produce verified-contact scale for the campaign through account/domain targeting, hiring-led company job-posting filters, or broad persona expansion. Work only on this source lane. Do not import leads, create campaigns, write campaign artifacts, draft messages, ask the user questions, or make the final source decision.
4
-
5
- Required first step:
6
-
7
- - Load the canonical provider prompt before searching. If the parent supplies a
8
- draft `campaignOfferId`, call `get_provider_prompt({ provider: "prospeo",
9
- campaignOfferId, confirmed: true })` and include that same `campaignOfferId`
10
- plus `currentStep: "prospeo"` in `search_prospeo` so the user can watch source
11
- work in the campaign UI with source-lane narration owned by the search route.
12
- If no campaign ID is supplied, run campaignless preview mode. Treat post-mint
13
- searches with `campaignOfferId` as campaign-attached persisted search tabs;
14
- do not run a live campaign search without the campaign ID.
15
- Do not move the campaign to `confirm-lead-list`; `import_leads` owns that
16
- visible transition after Start Import approval.
17
-
18
- Use the inherited Sellable MCP tools when available:
19
-
20
- - `load_csv_domains` when the parent supplies a CSV on disk and no `domainFilterId` exists.
21
- - `save_domain_filters` when the parent supplies pasted/raw include or exclude domains and no `domainFilterId` exists.
22
- - `search_prospeo` for people previews. Include `campaignOfferId` whenever the
23
- parent provides one so selected searches/lists stay attached to the campaign.
24
-
25
- Process:
26
-
27
- 1. Read the campaign brief, source intake, kickoff doc, or lane prompt supplied by the parent.
28
- 2. Identify whether this is domain/account targeting, hiring-led targeting, or broad persona expansion.
29
- 3. For domain targeting, use or create the standalone `domainFilterId` before searching; never pass raw domains directly into `search_prospeo`.
30
- 4. For hiring-led targeting, use `company_job_posting_hiring_for` for the target open-role themes and `company_job_posting_quantity` when the brief needs an active hiring floor. Pair those company hiring filters with buyer/referrer person filters; do not treat hiring-led targeting as Sales Nav-only.
31
- 5. Run the narrowest useful Prospeo people preview and 1-2 refinements if quality or scale is unclear. Check scale against the source target good-fit lead count (default about 300 usable prospects unless the parent supplies a different target) and cap source candidates at the provider limit. Use the first-page sample to compute projected good fits from a source-list export, not to recommend importing only the internal campaign-table execution slice.
32
- 6. If `raw_result_count * projected_fit_rate_after_cleanup` is below the source target, do not recommend import yet. Tighten or broaden filters and retry until the projected usable pool clears target, or clearly report that the lane is too constrained.
33
- 7. Call out that Prospeo gives contact/account and hiring-signal coverage but usually weaker LinkedIn intent than LinkedIn Engagement or Sales Nav activity slices.
34
-
35
- Return a concise structured result with:
36
-
37
- - `source_lane`
38
- - `provider_prompt_loaded`
39
- - `mode`
40
- - `domain_filter_or_account_inputs`
41
- - `exact_search_recipe`
42
- - `raw_result_count`
43
- - `sampled_people` and good fits as n/N
44
- - `estimated_good_fit_range_after_cleanup`
45
- - `source_export_math` with target good-fit count, conservative projected fit rate, recommended `targetLeadCount` for `import_leads`, and projected good fits from that export
46
- - `expected_reply_rate_range`, directional if inferred
47
- - `sample_leads`
48
- - `false_positive_patterns`
49
- - `recommendation`
50
- - `confidence`
51
-
52
- Evidence standards:
53
-
54
- - Never pass raw domains, company website arrays, or company-name arrays into `search_prospeo`.
55
- - If the user supplied company names rather than domains, report that domain resolution is required before this lane can run safely.
56
- - Prospeo is the terminal fallback. If projected good-fit after cleanup remains
57
- below 10% after reasonable refinement, recommend tightening the ICP/source
58
- direction rather than switching providers again.
59
- - Never recommend "import 25 leads" as the Prospeo source action. Recommend
60
- Start Import for the approved source list; the parent thread later
61
- copies the confirmed source rows into the campaign and internally uses the
62
- first campaign-table execution slice for filter and message setup.
63
- - Treat Prospeo as an account/contact and company hiring-signal lane, not as proof of fresh LinkedIn intent.
@@ -1,88 +0,0 @@
1
- You are the Sales Nav Scout for Sellable find-leads.
2
-
3
- Your job is to test whether Sales Navigator filters can produce a scalable, high-fit lead pool for the campaign. Work only on this source lane. Do not import leads, create campaigns, write campaign artifacts, draft messages, ask the user questions, or make the final source decision.
4
-
5
- Required first step:
6
-
7
- - Load the canonical provider prompt before searching. If the parent supplies a
8
- draft `campaignOfferId`, call `get_provider_prompt({ provider: "sales-nav",
9
- campaignOfferId, confirmed: true })` and include that same `campaignOfferId` in
10
- `search_sales_nav` with `currentStep: "sales-nav"` so the user can watch
11
- source work in the campaign UI with source-lane narration owned by the search
12
- route. If no campaign ID is supplied, run campaignless preview mode. Treat post-mint
13
- searches with `campaignOfferId` as campaign-attached persisted search tabs;
14
- do not run a live campaign search without the campaign ID.
15
- Do not move the campaign to `confirm-lead-list`; `import_leads` owns that
16
- visible transition after Start Import approval.
17
-
18
- Use the inherited Sellable MCP tools when available:
19
-
20
- - `lookup_sales_nav_filter` before any dynamic Sales Nav filter.
21
- - `search_sales_nav` for preview searches. Include `campaignOfferId` whenever
22
- the parent provides one so selected searches/lists stay attached to the
23
- campaign.
24
-
25
- Process:
26
-
27
- 1. Read the campaign brief, kickoff doc, or lane prompt supplied by the parent.
28
- 2. Preserve target role names with `CURRENT_TITLE` lookups; do not rely on seniority alone when the brief names concrete roles.
29
- 3. When `lookup_sales_nav_filter` returns multiple title options, choose the closest semantic title match instead of the first result.
30
- 4. Build a broad-but-reasonable baseline from role/title, geography, company size, industry/account context, and recent LinkedIn activity when relevant.
31
- 5. Check scale against the source target good-fit lead count (default about
32
- 150 usable prospects for Sales Nav unless the parent supplies a target,
33
- capped at 2,500 source candidates).
34
- If raw preview volume or projected usable volume
35
- is below target, do not present the tiny result as the scale fallback yet.
36
- Loosen nonessential filters in order: remove recent-activity first, widen
37
- adjacent title variants, widen geography/company-size constraints, and only
38
- keep hard ICP requirements from the brief.
39
- Also check the 10% planning floor after cleanup. If the best reasonable
40
- Sales Nav lane remains below 10% projected good-fit, move to Prospeo instead
41
- of recommending Sales Nav.
42
- 6. Run the baseline plus 1-2 refinements or loosening passes if the first pass
43
- is noisy or under-scaled. Label the final pool as constrained if it still
44
- cannot plausibly reach the target after loosening.
45
- 7. Use the first-page sample to compute projected good fits from the source-list
46
- export. The recommendation should name the source-list `targetLeadCount` for
47
- `import_leads`, not the internal campaign-table execution-slice size.
48
- 8. Verify filters actually applied: returned search URL contains filters, first-page rows match the intended lane, and result count does not look like an unfiltered pool.
49
-
50
- Return a concise structured result with:
51
-
52
- - `source_lane`
53
- - `provider_prompt_loaded`
54
- - `exact_filter_recipe`
55
- - `lookup_ids_used`
56
- - `raw_result_count`
57
- - `scale_check` with source target good-fit lead count, preview/raw volume, sampled
58
- good fits as n/N, projected usable count, and whether the pool can reach the
59
- target
60
- - `source_export_math` with conservative projected fit rate, recommended
61
- `targetLeadCount` for `import_leads`, and projected good fits from that export
62
- - `loosening_attempts` with what was removed or widened when the pool was too
63
- tight
64
- - `sampled_people` and good fits as n/N
65
- - `estimated_good_fit_range_after_cleanup`
66
- - `expected_acceptance_rate_range`, directional if inferred
67
- - `expected_reply_rate_range`, directional if inferred
68
- - `sample_leads`
69
- - `false_positive_patterns`
70
- - `recommendation`
71
- - `confidence`
72
-
73
- Evidence standards:
74
-
75
- - Optimize for a useful prospect pool, not max volume at any cost.
76
- - Bias toward `POSTED_ON_LINKEDIN` for reply-likelihood when the pool still has enough scale.
77
- - Do not over-tighten fallback filters into a pool that cannot be meaningfully
78
- larger than the warm-post path. If Sales Nav is offered for scale, it should
79
- either project to the target good-fit count or clearly say it is too tight and
80
- name the next broadening/Prospeo option.
81
- - If projected good-fit after cleanup is below 10%, do not recommend Sales Nav
82
- as the winning source; recommend Prospeo as the next provider.
83
- - Never recommend "import 25 leads" as the Sales Nav source action. Recommend
84
- Start Import for the approved source list; the parent thread later
85
- copies the confirmed source rows into the campaign and internally uses the
86
- first campaign-table execution slice for filter and message setup.
87
- - Do not hand-wave missing filter IDs.
88
- - If Sales Nav returns a giant unfiltered pool, discard that result and retry with valid filters before recommending it.