codex-orchestrator 2.0.17 → 2.0.19

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 (45) hide show
  1. package/CHANGELOG.md +22 -0
  2. package/README.md +40 -2
  3. package/dist/src/v2/acceptance-proof.d.ts.map +1 -1
  4. package/dist/src/v2/acceptance-proof.js +15 -5
  5. package/dist/src/v2/acceptance-proof.js.map +1 -1
  6. package/dist/src/v2/implementation-report.d.ts +3 -1
  7. package/dist/src/v2/implementation-report.d.ts.map +1 -1
  8. package/dist/src/v2/implementation-report.js +4 -1
  9. package/dist/src/v2/implementation-report.js.map +1 -1
  10. package/dist/src/v2/implementation-reviewer.d.ts +1 -0
  11. package/dist/src/v2/implementation-reviewer.d.ts.map +1 -1
  12. package/dist/src/v2/implementation-reviewer.js +8 -2
  13. package/dist/src/v2/implementation-reviewer.js.map +1 -1
  14. package/dist/src/v2/issue-check-policy.d.ts.map +1 -1
  15. package/dist/src/v2/issue-check-policy.js +25 -9
  16. package/dist/src/v2/issue-check-policy.js.map +1 -1
  17. package/dist/src/v2/run-issue.d.ts +1 -1
  18. package/dist/src/v2/run-issue.d.ts.map +1 -1
  19. package/dist/src/v2/run-issue.js +41 -31
  20. package/dist/src/v2/run-issue.js.map +1 -1
  21. package/dist/src/v2/runtime.d.ts.map +1 -1
  22. package/dist/src/v2/runtime.js +3 -2
  23. package/dist/src/v2/runtime.js.map +1 -1
  24. package/dist/src/v2/validation-progression.d.ts.map +1 -1
  25. package/dist/src/v2/validation-progression.js +3 -0
  26. package/dist/src/v2/validation-progression.js.map +1 -1
  27. package/internal-workflow/docs/agents/coding-skill-routing.md +4 -0
  28. package/internal-workflow/manifest.json +1 -1
  29. package/internal-workflow/profiles/spec_reviewer.toml +1 -1
  30. package/internal-workflow/profiles/standards_reviewer.toml +1 -1
  31. package/internal-workflow/skills/code-review/SKILL.md +5 -0
  32. package/internal-workflow/skills/implement/SKILL.md +11 -9
  33. package/internal-workflow/skills/implement/evals/evals.json +0 -12
  34. package/internal-workflow/skills/plan/SKILL.md +32 -12
  35. package/internal-workflow/skills/plan/agents/openai.yaml +2 -2
  36. package/internal-workflow/skills/plan/evals/evals.json +27 -0
  37. package/internal-workflow/skills/plan-review/SKILL.md +110 -0
  38. package/internal-workflow/skills/plan-review/agents/openai.yaml +6 -0
  39. package/internal-workflow/skills/plan-review/evals/evals.json +57 -0
  40. package/internal-workflow/skills/tickets-orchestrator/SKILL.md +0 -4
  41. package/internal-workflow/skills/to-spec/SKILL.md +4 -2
  42. package/internal-workflow/skills/to-tickets/SKILL.md +30 -24
  43. package/internal-workflow/skills/to-tickets/agents/openai.yaml +1 -1
  44. package/internal-workflow/skills/to-tickets/evals/evals.json +30 -5
  45. package/package.json +1 -1
@@ -1,6 +1,6 @@
1
1
  ---
2
2
  name: to-tickets
3
- description: Compile Plan's contextual PRD draft or approved product authority into the smallest executable ticket packet, run fresh semantic review per settled revision, obtain one approval, publish serially, verify tracker read-back, and stop before implementation.
3
+ description: Compile Plan's contextual PRD draft or approved product authority into the smallest executable ticket packet, run complete semantic review with one fresh final review after any consolidated repair, obtain one approval, publish serially, verify tracker read-back, and stop before implementation.
4
4
  ---
5
5
 
6
6
  # To Tickets
@@ -15,8 +15,9 @@ ticket contains only its local observable outcome, scope, blockers, proof, and
15
15
  Parent link. The approved ticket body plus Parent PRD is the final execution
16
16
  authority; no later planning pass is inserted.
17
17
 
18
- `$to-tickets` owns one fresh reviewer per settled revision, one explicit user
19
- approval, serialized publication, reconciliation, and tracker read-back.
18
+ `$to-tickets` owns complete whole-packet `$plan-review`, one fresh final review
19
+ after any consolidated repair, one explicit user approval, serialized
20
+ publication, reconciliation, and tracker read-back.
20
21
 
21
22
  The primary route remains `planning-only` while compiling or publishing these
22
23
  artifacts. `ticket-graph` begins only after separate delivery authorization;
@@ -45,10 +46,6 @@ outcome and blocking relationship. Risk strengthens proof and review focus; it
45
46
  does not create tickets. Merge steps that share one owner, release, and
46
47
  validation path.
47
48
 
48
- Do not publish a child whose outcome is only transient evidence or a tracker
49
- comment. Resolve that discovery before compiling the executable graph, or make
50
- the required durable result a committed artifact.
51
-
52
49
  Where repository evidence proves a small prefactor is required, "Make the
53
50
  change easy, then make the easy change." Keep it inside the same vertical
54
51
  outcome unless it is independently verifiable and genuinely blocks later work.
@@ -144,12 +141,12 @@ an explicit blocking discovery outcome. The expected review verdict must be
144
141
 
145
142
  ### 4. Run one fresh semantic packet review
146
143
 
147
- For each settled revision composed by `$to-tickets` before approval or
148
- publication, launch exactly one fresh `standards_reviewer` without history
144
+ After the complete packet satisfies the freeze conditions, invoke
145
+ `$plan-review` with exactly one fresh `plan_reviewer` without history
149
146
  (`fork_context=false` on V1; `fork_turns="none"` on V2). Begin with
150
- `Assigned role: standards_reviewer`, provide the full authority and packet, and
151
- apply all three lenses in the same activation: source fidelity, ticket
152
- executability, and minimum solution and scope conservation.
147
+ `Assigned role: plan_reviewer`, provide the full authority and packet, and apply
148
+ all three lenses in the same activation: source fidelity, ticket executability,
149
+ and minimum solution and scope conservation.
153
150
 
154
151
  - **Source fidelity:** every product claim, decision, scope boundary,
155
152
  consequence, and blocker is grounded in the source; no behavior or ownership
@@ -185,19 +182,28 @@ Before verdict, traverse each applicable seam:
185
182
  Return concise evidence for applicable seams and omit `N/A` seams. This is
186
183
  review output, not another packet artifact.
187
184
 
188
- The reviewer is read-only and returns `APPROVE` or `NEEDS_WORK` with exact
189
- source/artifact evidence. Capture a non-empty fresh child identity and wait for
190
- that same child. Do not split the lenses across children and do not launch the
191
- delivery Standards review. After `NEEDS_WORK`, repair the packet and launch a
192
- new reviewer without prior review history over the complete revised packet. A
193
- same-context or findings-only recheck cannot approve. Repeat until the latest
194
- revision receives one full approval. Missing or failed review blocks
195
- publication; product-changing repairs return to the user.
196
-
197
- Before re-review, repair every verified blocker and sweep the complete packet
198
- for the same defect class. If a blocker exposes unfinished inventory,
185
+ The reviewer is read-only and returns `APPROVE` or `NEEDS_WORK` with every
186
+ material blocker found after completing all applicable lenses and seam
187
+ traversals, plus exact source/artifact evidence. It must complete the
188
+ whole-packet review and must not stop after representative findings. Capture a
189
+ non-empty fresh child identity and wait for that same child. Do not split the
190
+ lenses across children and do not launch the delivery Standards review.
191
+
192
+ When review returns a plan-changing `QUESTION`, return it through Plan to the
193
+ user and stop without revising or publishing the packet. Resume only after the
194
+ answer becomes authority and the complete packet is settled again.
195
+
196
+ After `NEEDS_WORK` with verified blockers, repair all blockers together and
197
+ sweep the complete packet for every defect class they expose. Do not review
198
+ intermediate repair edits. If a blocker exposes unfinished inventory,
199
199
  authority, or composition, return to drafting and pass the freeze conditions
200
- again.
200
+ again. Once the complete revised packet satisfies the freeze conditions,
201
+ launch one fresh final whole-packet `plan_reviewer` without prior review
202
+ history. Use one fresh reviewer per settled revision; a same-context recheck
203
+ cannot approve. Further review is justified only by a remaining material
204
+ blocker; optional hardening, broader proof, stylistic refinement, or
205
+ requirements without authority do not reopen the review cycle. Missing or
206
+ failed review blocks publication; product-changing repairs return to the user.
201
207
 
202
208
  ### 5. Get one explicit approval
203
209
 
@@ -1,6 +1,6 @@
1
1
  interface:
2
2
  display_name: "To Tickets"
3
3
  short_description: "Review and publish executable ticket packets"
4
- default_prompt: "Use $to-tickets to compile and fully review each settled packet revision, obtain one approval, publish serially, verify tracker read-back, and stop before implementation."
4
+ default_prompt: "Use $to-tickets to compile a complete packet, review it fully, consolidate any repairs before one fresh final review, obtain one approval, publish serially, verify tracker read-back, and stop before implementation."
5
5
  policy:
6
6
  allow_implicit_invocation: false
@@ -15,10 +15,10 @@
15
15
  "forbidden": ["create duplicate durable planning state", "request another execution artifact"]
16
16
  },
17
17
  {
18
- "id": "complete-packet-gets-one-review-per-revision",
18
+ "id": "complete-packet-review-after-freeze",
19
19
  "prompt": "Prepare a publish-ready PRD and three-ticket graph after Plan passed the combined PRD draft to you.",
20
- "expected": ["launch one fresh semantic reviewer over each settled complete revision", "apply source fidelity and ticket executability in that review", "obtain one explicit user approval before publication"],
21
- "forbidden": ["review the intermediate PRD", "launch separate workflow owners for each lens", "request multiple approvals"]
20
+ "expected": ["launch one fresh semantic reviewer only after the complete packet satisfies the freeze conditions", "apply source fidelity and ticket executability in that review", "obtain one explicit user approval before publication"],
21
+ "forbidden": ["review the intermediate PRD", "review intermediate repair edits", "launch separate workflow owners for each lens", "request multiple approvals"]
22
22
  },
23
23
  {
24
24
  "id": "serialized-publication-recovers-response-loss",
@@ -102,8 +102,8 @@
102
102
  {
103
103
  "id": "repair-gets-fresh-full-packet-review",
104
104
  "prompt": "Semantic review returns NEEDS_WORK and the packet is repaired before user approval.",
105
- "expected": ["repair every verified blocker and sweep the complete packet for the same defect class", "pass the settled packet conditions again", "launch one new reviewer without prior review history", "review the complete revised packet", "keep one user approval for the final packet"],
106
- "forbidden": ["continue with the same reviewer", "approve from a findings-only recheck", "launch multiple reviewers for one revision"]
105
+ "expected": ["repair all verified blockers together and sweep the complete packet for every defect class they expose", "pass the freeze conditions again", "launch one fresh final reviewer without prior review history", "review the complete revised packet", "keep one user approval for the final packet"],
106
+ "forbidden": ["review intermediate repair edits", "continue with the same reviewer", "approve from a findings-only recheck", "reopen review for optional hardening or requirements without authority"]
107
107
  },
108
108
  {
109
109
  "id": "private-design-stays-with-implement",
@@ -128,6 +128,31 @@
128
128
  "prompt": "UI acceptance requires dynamic content absent from the proposed API or DTO.",
129
129
  "expected": ["reject the packet until an existing or authorized data contract supplies the content"],
130
130
  "forbidden": ["let the UI invent the data", "add a new owner solely for proof"]
131
+ },
132
+ {
133
+ "id": "final-packet-uses-plan-review-once",
134
+ "prompt": "Compile an approved PRD into a complete publish-ready Parent and multi-ticket packet.",
135
+ "expected": [
136
+ "do not review the intermediate PRD",
137
+ "invoke one fresh plan-review over the complete frozen packet",
138
+ "cover source fidelity, ticket executability, and minimum solution in that review"
139
+ ],
140
+ "forbidden": [
141
+ "invoke standards_reviewer for planning review",
142
+ "review both the PRD and final packet"
143
+ ]
144
+ },
145
+ {
146
+ "id": "plan-review-question-returns-to-user",
147
+ "prompt": "The final packet review raises an unresolved ownership question but finds no verified blocker.",
148
+ "expected": [
149
+ "return the question through Plan to the user",
150
+ "resume only after the answer becomes authority and the packet is settled again"
151
+ ],
152
+ "forbidden": [
153
+ "guess the answer",
154
+ "repair or publish the packet"
155
+ ]
131
156
  }
132
157
  ]
133
158
  }
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "codex-orchestrator",
3
- "version": "2.0.17",
3
+ "version": "2.0.19",
4
4
  "description": "Reusable GitHub Issues runner for Codex.",
5
5
  "type": "module",
6
6
  "license": "MIT",