clawdi 0.14.44 → 0.14.46

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.
Files changed (4) hide show
  1. package/dist/index.js +191 -191
  2. package/package.json +1 -1
  3. package/skills/clawdi/SKILL.md +83 -81
  4. package/skills/hosted-versions/1/clawdi/SKILL.md +79 -31
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "clawdi",
3
- "version": "0.14.44",
3
+ "version": "0.14.46",
4
4
  "description": "The best home for all your AI agents. Run them in the cloud or connect your own—with their context and tools in one place.",
5
5
  "license": "MIT",
6
6
  "type": "module",
package/skills/clawdi/SKILL.md CHANGED
@@ -1,87 +1,58 @@
1
1
  ---
2
2
  name: clawdi
3
- description: "Cross-agent long-term memory + session history for the current user: their preferences, coding habits, named projects / repos / tools, past bugs and architecture decisions, AND their past agent conversations across Claude Code / Codex / OpenClaw / Hermes. Surface this skill BEFORE answering any question about the user themselves, their work, or their history — even when phrased abstractly (e.g. 'what do I usually use for X', 'find the session where I worked on auth'). Also provides connected-service tools (Gmail, GitHub, Notion, Drive, Calendar, etc.) and reads Clawdi share URLs (https://cloud.clawdi.ai/s/...) the user pastes."
3
+ description: "Use Clawdi Cloud for missing user memory, past sessions, Project or Vault context, Clawdi share URLs, and connected-service fallback such as Gmail, GitHub, Notion, Drive, or Calendar. Prefer an authenticated official service CLI; otherwise choose a trusted direct MCP, safely installable official CLI, official API or SDK, or the Clawdi connector as fallback. Do not invoke solely because a project, person, repo, or tool is named."
4
4
  ---
5
5
 
6
6
  # Clawdi Cloud
7
7
 
8
- You have access to Clawdi Cloud tools via the `clawdi` MCP server. Use them aggressively — memory + session retrieval is the highest-leverage capability you have here.
8
+ Use Clawdi Cloud tools through the `clawdi` MCP server when they provide context or
9
+ capabilities unavailable more directly.
9
10
 
10
- ## Memory
11
-
12
- Three tools for cross-agent memory:
13
-
14
- - `memory_search` — Search long-term memory by natural-language query (any language).
15
- - `memory_add` — Save a durable memory for cross-agent recall. Do not store plaintext tokens, API keys, or bearer credentials; store those in Vault and save only the `clawdi://` reference. Categories: `fact` (technical facts, API details, config values), `preference` (user preferences, coding style, workflow choices), `pattern` (recurring patterns, pitfalls, team conventions), `decision` (architecture decisions and their reasoning), `context` (project context, deadlines, ongoing work).
16
- - `memory_extract` — Batch-extract durable memories from the CURRENT conversation. Call this when the user says "extract memories", "save what we discussed", "remember this conversation", or equivalent. The tool returns instructions that walk you through a list-then-confirm flow using `memory_search` and `memory_add` — follow them exactly, including **waiting for the user's approval before writing anything**. Never skip the confirmation step, never save more than 5 memories in one invocation, and do not narrate your internal workflow to the user.
17
-
18
- ### When to search — bias toward calling
19
-
20
- **Default assumption: the user has stored context you don't have. Call `memory_search` BEFORE answering any question about them, their project, their preferences, or their history. A call that returns empty costs ~100ms; a missed hit makes you look amnesic and forces them to re-teach you every session.**
21
-
22
- The single most common failure mode is NOT calling memory_search on abstract self-referential questions. If the user's message has any of these shapes, you MUST call it — no judgment, no exceptions:
23
-
24
- 1. **Preference / habit questions**, even without a specific entity named.
25
- Examples: "what do I usually use for X", "how do I normally do Y", "what's my preferred tool for Z", "what's my coding style". Pass a short paraphrase as the query.
26
- 2. **Callbacks to prior context.** "as I mentioned", "like last time", "you know the one", "we discussed before", "what was that X we set up".
27
- 3. **Named entities specific to this user.** Their project / repo / service / team / tool name. A person by name.
28
- 4. **Past bugs, decisions, investigations, design choices.**
29
- 5. **Start of a new session where they reference anything about themselves or their work.**
11
+ ## Context Routing
30
12
 
31
- Do NOT search for:
32
- - Purely textbook programming questions with no user-specific signal ("how does `useEffect` work", "what is the time complexity of quicksort").
33
- - Questions the current code already answers directly.
13
+ Use the current conversation and user-provided artifacts first. For project facts, inspect
14
+ the workspace, repository documentation, and local history. Use `memory_search` only for
15
+ missing user-specific preferences, decisions, or prior context. Use `session_list`,
16
+ `session_search`, and `session_get` only when the user asks for past conversations or
17
+ transcript-level detail is necessary. Do not call Memory and Session speculatively or in parallel.
18
+ A named entity alone does not justify a Cloud lookup, and an empty Memory result does not justify
19
+ a Session search.
34
20
 
35
- **When unsure, search.** Empty results cost you nothing. Missing the user's context costs you their trust.
36
-
37
- ### When to save
38
-
39
- - After fixing a non-obvious bug (save root cause + fix)
40
- - After making an architecture decision (save reasoning)
41
- - After discovering a useful pattern or workaround
42
- - When the user explicitly says "remember this" / "save this"
43
- - After learning a user preference you'd otherwise have to re-ask ("I prefer rg", "I always use pnpm")
21
+ ## Memory
44
22
 
45
- Write memories as standalone sentences with full context — include names, not pronouns. A future session will read this without knowing today's conversation.
23
+ Memory is durable user-specific context shared across agents.
46
24
 
47
- Do NOT save trivial facts that are obvious from the code itself, or generic programming knowledge.
25
+ - `memory_search` — Search durable memory by natural-language query.
26
+ - `memory_list` — Review stored memories and their stable IDs.
27
+ - `memory_create` — Save a durable fact, preference, pattern, decision, or project context.
28
+ - `memory_update` — Replace one exact memory's content without changing its metadata.
29
+ - `memory_delete` — Delete one exact memory by ID.
30
+ - `memory_extract` — Prepare memories from the current conversation. Follow its returned
31
+ review-and-confirm instructions and wait for user approval before calling `memory_create`.
48
32
 
49
- Do NOT save plaintext tokens, API keys, bearer credentials, or private keys in memory. Use Vault for secret values and save only a `clawdi://` reference when future agents need to know where a secret lives.
33
+ Use `memory_create` for explicit "remember this" requests or durable user-specific preferences
34
+ and decisions not discoverable from the repository. Ask when persistence is unclear. Do not
35
+ save routine task completion, code facts, speculation, or plaintext secrets; use Vault and
36
+ remember only the exact `clawdi://` reference. List before updating or deleting unless the user
37
+ already supplied the exact memory ID; never infer which stored item to mutate.
50
38
 
51
39
  ## Sessions
52
40
 
53
- Two tools for reading and finding past agent conversations stored in Clawdi Cloud:
54
-
55
- - `session_read` — Fetch a single session by reference and return its full conversation as Markdown. Accepts a Clawdi share URL (`https://cloud.clawdi.ai/s/{uuid}`) OR a session UUID for one of the user's own sessions. Handles owned and shared sessions transparently — you don't need to know which one.
56
- - `session_search` — Find sessions in the user's history by keyword. Trigram-ranked substring search with typo tolerance. Returns matching sessions with summary, project, timestamps, and **session UUIDs you can pass back to `session_read`**.
57
-
58
- ### When to read — call `session_read` whenever the user references a specific session
59
-
60
- MUST call when the user's message includes:
61
- - A Clawdi share URL (e.g. `https://cloud.clawdi.ai/s/11111111-2222-3333-4444-555555555555`) — pass the full URL
62
- - A direct reference like "open the session where I did X" or "the one from yesterday about auth" — first call `session_search` to find the UUID, then `session_read` to load it
41
+ - Use `session_list` to browse recent sessions or filter by time, Agent, or Project.
42
+ - Use `session_search` to find past agent conversations by keyword and obtain session UUIDs.
43
+ - Use `session_get` to read a session by UUID or Clawdi share URL.
63
44
 
64
- Do NOT call WebFetch on `cloud.clawdi.ai/s/...` URLs — `session_read` is the right tool and avoids the WebFetch permission prompt.
45
+ Call `session_get` when the user provides a Clawdi share URL or session UUID and wants its
46
+ contents. For a request to open a specific unnamed past conversation, use `session_search`
47
+ to find the UUID and then read the selected match.
65
48
 
66
- ### When to search — bias toward calling, similar to `memory_search`
67
-
68
- MUST call `session_search` when:
69
- - The user asks about prior work: "what did I do about the focus bug", "find the session where I migrated auth", "show me last week's debugging session"
70
- - They reference a past investigation by topic but don't name a specific session
71
- - They want to continue / reuse approach from a prior conversation
72
-
73
- ### Difference from `memory_search`
74
-
75
- - `memory_search` finds **stored facts / preferences / decisions** the user (or a previous agent run) explicitly extracted. Short rows; high signal.
76
- - `session_search` finds **full conversations** in the corpus. Long rows; useful when the user wants the original context, not just the takeaway.
77
-
78
- When the user's request is **conceptual** ("how do I usually do X"), prefer `memory_search`. When they want to **revisit a specific past conversation** ("the session where..."), use `session_search`. When unsure, try `memory_search` first (cheaper, faster), fall back to `session_search` if empty.
49
+ Do NOT call WebFetch on `cloud.clawdi.ai/s/...` URLs — `session_get` is the right tool and avoids the WebFetch permission prompt.
79
50
 
80
51
  ## Projects
81
52
 
82
53
  Three read-only tools expose the caller's visible Project context:
83
54
 
84
- - `project_current` — Read the current or runtime-bound Project.
55
+ - `project_current_get` — Read the current or runtime-bound Project.
85
56
  - `project_list` — List visible Projects.
86
57
  - `project_get` — Read one visible Project by UUID.
87
58
 
@@ -89,9 +60,9 @@ Hosted runtimes see only their bound Project. Treat a not-found response as an
89
60
  access boundary as well as a possible unknown UUID; do not try to bypass it
90
61
  through another tool.
91
62
 
92
- ## Vault Metadata
63
+ ## Vault
93
64
 
94
- Two read-only MCP tools expose safe Vault metadata without secret values:
65
+ Vault read tools expose metadata and exact references:
95
66
 
96
67
  - `vault_list` — List Vault attachments and key counts for visible Projects.
97
68
  - `vault_get` — List key names, provenance, and exact `clawdi://` references for one attached Vault.
@@ -110,6 +81,19 @@ to an authorized runtime:
110
81
  Use the live schemas from the `clawdi` MCP server as authoritative; the local
111
82
  stdio command only transports the protocol.
112
83
 
84
+ Vault write tools are available for explicit user requests:
85
+
86
+ - `vault_create` — Create a Vault attached to one exact owner Project.
87
+ - `vault_item_upsert` — Create or replace exact fields in an attached Vault.
88
+ - `vault_item_delete` — Delete exact fields from a single-Project Vault.
89
+
90
+ Follow the live schema and supply every required Project, Vault, section, and field identity;
91
+ never infer an overwrite or deletion. Treat field values as sensitive inputs and never echo
92
+ them, save them to Memory, or include them in logs. Environment-bound callers may write only
93
+ their bound Project, and field deletion is rejected when a Vault is attached to multiple
94
+ Projects. Whole-Vault deletion, attach/detach, bulk import, and credential profiles remain
95
+ foreground operator workflows; never bypass that boundary through raw HTTP or daemon RPC.
96
+
113
97
  ## Wallet Funding
114
98
 
115
99
  Use `clawdi wallet status --json` to inspect the authenticated Wallet balance, verified
@@ -118,12 +102,39 @@ browser wallet surface; Clawdi does not store the payment private key. Ask the u
118
102
  there. Command-line spending requires a future owner-only or hardware signer authority and is not
119
103
  available.
120
104
 
121
- ## Connectors
105
+ ## Connector Routing
106
+
107
+ Respect an explicit user choice. Otherwise inspect installed service CLIs, direct MCP tools
108
+ already exposed by the runtime, and authorized API or SDK credentials. If an installed and
109
+ authenticated official CLI can perform the task, use it directly. Check availability and
110
+ authentication non-destructively and prefer structured output.
111
+
112
+ Otherwise choose the lowest-setup reliable option for the task. Consult the service's official
113
+ documentation when installation, authentication, commands, or schemas are uncertain or likely
114
+ to have changed:
115
+
116
+ - Use a trusted direct MCP already configured and exposed by the runtime. Do not automatically
117
+ download, install, or start an unfamiliar MCP server.
118
+ - Safely install the official CLI when the runtime permits it, the source is verified as
119
+ official, and no elevation or persistent host change is required.
120
+ - Use the official API or SDK with a verified contract and credentials already authorized for
121
+ the runtime, including through an exact Vault reference.
122
+ - Use the Clawdi connector when no direct option can perform the operation.
123
+
124
+ Before a side effect, establish the exact service account and organization, Project, or tenant.
125
+ Use `connector_account_list` when the connector is a candidate and its identity is not already clear.
126
+ Fallback must not silently change that identity. Do not scan for credentials, start an interactive
127
+ login, invent API details, or expose secrets. Choose the path before a side effect and advance
128
+ only after a definite preflight failure. If a mutation's result is ambiguous, inspect it through
129
+ the same path; never repeat it through another path.
130
+
131
+ ## Connector Workflow
122
132
 
123
- Use the Composio Tool Router meta-tools returned by `tools/list` on the `clawdi` MCP server.
124
- Treat their live names and schemas as authoritative; never assume a fixed meta-tool set.
133
+ When the Clawdi connector path is selected, use the Composio Tool Router meta-tools returned
134
+ by `tools/list` on the `clawdi` MCP server. Treat their live names and schemas as
135
+ authoritative; never assume a fixed meta-tool set.
125
136
 
126
- 1. Start each external-app workflow with `COMPOSIO_SEARCH_TOOLS`. Follow its exposed
137
+ 1. Start the connector workflow with `COMPOSIO_SEARCH_TOOLS`. Follow its exposed
127
138
  `queries` and `session` schema, reuse the returned session ID throughout that workflow,
128
139
  and use only the exact toolkit and tool slugs it returns. If a required schema is absent
129
140
  or incomplete, call `COMPOSIO_GET_TOOL_SCHEMAS`; never invent fields or inputs.
@@ -148,19 +159,10 @@ Treat their live names and schemas as authoritative; never assume a fixed meta-t
148
159
  arguments. Batch only independent calls. Keep ordinary results inline. Set
149
160
  `sync_response_to_workbench` only when a result may be large or needs later remote
150
161
  processing; use `COMPOSIO_REMOTE_WORKBENCH` / `COMPOSIO_REMOTE_BASH_TOOL` only for large
151
- responses saved remotely or remote artifacts.
152
- 6. Preserve dependencies and returned semantics. Follow signed-file metadata, pagination
153
- fields, and termination signals exactly as exposed. Select an account only when the schema
154
- supports it, and use additional or future meta-tools only according to their live schemas.
155
-
156
- ## Vault Management
157
-
158
- Vault mutation is intentionally not an Agent MCP capability. Do not use raw HTTP, daemon
159
- control RPC, or execute foreground Vault CLI commands on the user's behalf. When the user
160
- asks to write, import, attach, detach, or delete Vault data, explain that a human operator
161
- must perform it and provide the safest exact foreground command. Prefer `clawdi vault set
162
- KEY --prompt` for one value and `clawdi vault import ...` for migrations; never place a
163
- plaintext secret in command arguments or your response.
162
+ responses saved remotely or remote artifacts. Preserve dependencies and returned semantics;
163
+ follow signed-file metadata, pagination fields, and termination signals exactly as exposed.
164
+ Select an account only when the schema supports it, and use additional or future meta-tools
165
+ only according to their live schemas.
164
166
 
165
167
  ## AI Provider Management
166
168
 
package/skills/hosted-versions/1/clawdi/SKILL.md CHANGED
@@ -1,47 +1,66 @@
1
1
  ---
2
2
  name: clawdi
3
- description: "Use the user's account-wide long-term memory and past agent sessions; inspect the current Hosted Project and safe Vault metadata; use connected services such as Gmail, GitHub, Notion, Drive, and Calendar; and read Clawdi share URLs."
3
+ description: "Use Clawdi Cloud for missing user memory, past sessions, Project or Vault context, Clawdi share URLs, and connected-service fallback such as Gmail, GitHub, Notion, Drive, or Calendar. Prefer an authenticated official service CLI; otherwise choose a trusted direct MCP, safely installable official CLI, official API or SDK, or the Clawdi connector as fallback. Do not invoke solely because a project, person, repo, or tool is named."
4
4
  ---
5
5
 
6
6
  # Clawdi Cloud
7
7
 
8
8
  Use Clawdi Cloud tools through the `clawdi` MCP server. Treat the live tool schemas as authoritative.
9
9
 
10
+ ## Hosted Boundary
11
+
12
+ Third-party tool routing below applies unchanged in Hosted. Do not inspect, run, or suggest
13
+ Clawdi host-management commands such as `clawdi setup`, `clawdi wallet`, `clawdi vault`, or
14
+ `clawdi ai-provider`.
15
+
16
+ ## Context Routing
17
+
18
+ Use the current conversation and user-provided artifacts first. For project facts, inspect
19
+ the workspace, repository documentation, and local history. Use `memory_search` only for
20
+ missing user-specific preferences, decisions, or prior context. Use `session_list`,
21
+ `session_search`, and `session_get` only when the user asks for past conversations or
22
+ transcript-level detail is necessary. Do not call Memory and Session speculatively or in parallel.
23
+ A named entity alone does not justify a Cloud lookup, and an empty Memory result does not justify
24
+ a Session search.
25
+
10
26
  ## Memory
11
27
 
12
28
  Memory is shared across the user's Hosted agents, not isolated to the current agent.
13
29
 
14
30
  - `memory_search` — Search durable memory by natural-language query.
15
- - `memory_add` — Save a durable fact, preference, pattern, decision, or project context.
31
+ - `memory_list` — Review stored memories and their stable IDs.
32
+ - `memory_create` — Save a durable fact, preference, pattern, decision, or project context.
33
+ - `memory_update` — Replace one exact memory's content without changing its metadata.
34
+ - `memory_delete` — Delete one exact memory by ID.
16
35
  - `memory_extract` — Prepare memories from the current conversation. Follow its returned
17
- review-and-confirm instructions and wait for user approval before calling `memory_add`.
18
-
19
- Search before answering questions about the user's preferences, projects, prior decisions,
20
- recurring workflows, or earlier bugs. Save useful non-obvious outcomes and explicit
21
- "remember this" requests as standalone statements with enough context for another agent.
36
+ review-and-confirm instructions and wait for user approval before calling `memory_create`.
22
37
 
23
- Never store plaintext tokens, API keys, bearer credentials, or private keys in memory. Store
24
- secrets in Vault and remember only an exact `clawdi://` reference when useful.
38
+ Use `memory_create` for explicit "remember this" requests or durable user-specific preferences
39
+ and decisions not discoverable from the repository. Ask when persistence is unclear. Do not
40
+ save routine task completion, code facts, speculation, or plaintext secrets; use Vault and
41
+ remember only the exact `clawdi://` reference. List before updating or deleting unless the user
42
+ already supplied the exact memory ID; never infer which stored item to mutate.
25
43
 
26
44
  ## Sessions
27
45
 
46
+ - Use `session_list` to browse recent sessions or filter by time, Agent, or Project.
28
47
  - Use `session_search` to find past agent conversations by keyword and obtain session UUIDs.
29
- - Use `session_read` to read a session by UUID or Clawdi share URL.
48
+ - Use `session_get` to read a session by UUID or Clawdi share URL.
30
49
 
31
- Call `session_read` when the user provides a Clawdi share URL. When the user refers to a
32
- past conversation without a UUID, call `session_search` first and then read the matching
33
- session. Do not use a generic web fetcher for Clawdi share URLs.
50
+ Call `session_get` when the user provides a Clawdi share URL or session UUID and wants its
51
+ contents. Use `session_search` to locate a requested unnamed conversation. Do not use a
52
+ generic web fetcher for Clawdi share URLs.
34
53
 
35
54
  ## Projects
36
55
 
37
- - `project_current` — Read the runtime-bound Project.
56
+ - `project_current_get` — Read the runtime-bound Project.
38
57
  - `project_list` — List Projects visible to the caller.
39
58
  - `project_get` — Read one visible Project by UUID.
40
59
 
41
60
  A Hosted runtime is restricted to its bound Project. Treat not-found as an access boundary
42
61
  as well as a possible unknown UUID; do not try to bypass it through another tool.
43
62
 
44
- ## Vault Metadata
63
+ ## Vault
45
64
 
46
65
  - `vault_list` — List attached Vaults and key counts for visible Projects.
47
66
  - `vault_get` — List key names, provenance, and exact references for an attached Vault.
@@ -56,23 +75,52 @@ The metadata tools never return plaintext secret values. Preserve exact referenc
56
75
  - `clawdi://project/<project-id>/vault/<vault>/field/<field>`
57
76
  - `clawdi://project/<project-id>/vault/<vault>/section/<section>/field/<field>`
58
77
 
59
- Vault mutation is not an Agent MCP capability. Ask the user to manage Vault data through an
60
- authorized human-facing surface; never call raw HTTP or invent an unavailable tool.
78
+ Vault write tools are available for explicit user requests:
79
+
80
+ - `vault_create` — Create a Vault attached to the runtime-bound Project.
81
+ - `vault_item_upsert` — Create or replace exact fields in an attached Vault.
82
+ - `vault_item_delete` — Delete exact fields from a single-Project Vault.
83
+
84
+ Follow the live schema and supply every required Project, Vault, section, and field identity;
85
+ never infer an overwrite or deletion. Treat field values as sensitive inputs and never echo
86
+ them, save them to Memory, or include them in logs. Hosted writes are restricted to the
87
+ runtime-bound Project, and field deletion is rejected when a Vault is attached to multiple
88
+ Projects. Whole-Vault deletion, attach/detach, bulk import, and credential profiles remain
89
+ unavailable through Agent MCP; do not bypass that boundary through raw HTTP.
90
+
91
+ ## Connector Routing
92
+
93
+ Respect an explicit user choice. Otherwise inspect installed service CLIs, direct MCP tools
94
+ already exposed by the runtime, and authorized API or SDK credentials. If an installed and
95
+ authenticated official CLI can perform the task, use it directly. Check availability and
96
+ authentication non-destructively and prefer structured output.
97
+
98
+ Otherwise choose the lowest-setup reliable option for the task. Consult the service's official
99
+ documentation when installation, authentication, commands, or schemas are uncertain or likely
100
+ to have changed:
61
101
 
62
- ## Wallet Funding
102
+ - Use a trusted direct MCP already configured and exposed by the runtime. Do not automatically
103
+ download, install, or start an unfamiliar MCP server.
104
+ - Safely install the official CLI when the runtime permits it, the source is verified as
105
+ official, and no elevation or persistent host change is required.
106
+ - Use the official API or SDK with a verified contract and credentials already authorized for
107
+ the runtime, including through an exact Vault reference.
108
+ - Use the Clawdi connector when no direct option can perform the operation.
63
109
 
64
- Use `clawdi wallet status --json` to inspect the authenticated Wallet balance, verified
65
- binding, and x402 readiness. Binding and Base USDC top-up are available only through the
66
- browser wallet surface; Clawdi does not store the payment private key. Ask the user to fund
67
- there. Command-line spending requires a future owner-only or hardware signer authority and is not
68
- available.
110
+ Before a side effect, establish the exact service account and organization, Project, or tenant.
111
+ Use `connector_account_list` when the connector is a candidate and its identity is not already clear.
112
+ Fallback must not silently change that identity. Do not scan for credentials, start an interactive
113
+ login, invent API details, or expose secrets. Choose the path before a side effect and advance
114
+ only after a definite preflight failure. If a mutation's result is ambiguous, inspect it through
115
+ the same path; never repeat it through another path.
69
116
 
70
- ## Connectors
117
+ ## Connector Workflow
71
118
 
72
- Use the Composio Tool Router meta-tools returned by `tools/list` on the `clawdi` MCP server.
73
- Treat their live names and schemas as authoritative; never assume a fixed meta-tool set.
119
+ When the Clawdi connector path is selected, use the Composio Tool Router meta-tools returned
120
+ by `tools/list` on the `clawdi` MCP server. Treat their live names and schemas as
121
+ authoritative; never assume a fixed meta-tool set.
74
122
 
75
- 1. Start each external-app workflow with `COMPOSIO_SEARCH_TOOLS`. Follow its exposed
123
+ 1. Start the connector workflow with `COMPOSIO_SEARCH_TOOLS`. Follow its exposed
76
124
  `queries` and `session` schema, reuse the returned session ID throughout that workflow,
77
125
  and use only the exact toolkit and tool slugs it returns. If a required schema is absent
78
126
  or incomplete, call `COMPOSIO_GET_TOOL_SCHEMAS`; never invent fields or inputs.
@@ -97,7 +145,7 @@ Treat their live names and schemas as authoritative; never assume a fixed meta-t
97
145
  arguments. Batch only independent calls. Keep ordinary results inline. Set
98
146
  `sync_response_to_workbench` only when a result may be large or needs later remote
99
147
  processing; use `COMPOSIO_REMOTE_WORKBENCH` / `COMPOSIO_REMOTE_BASH_TOOL` only for large
100
- responses saved remotely or remote artifacts.
101
- 6. Preserve dependencies and returned semantics. Follow signed-file metadata, pagination
102
- fields, and termination signals exactly as exposed. Select an account only when the schema
103
- supports it, and use additional or future meta-tools only according to their live schemas.
148
+ responses saved remotely or remote artifacts. Preserve dependencies and returned semantics;
149
+ follow signed-file metadata, pagination fields, and termination signals exactly as exposed.
150
+ Select an account only when the schema supports it, and use additional or future meta-tools
151
+ only according to their live schemas.