feat(skills): add shipping-features orchestrator + capturing-knowledge#9873
Open
pa-lem wants to merge 18 commits into
Open
feat(skills): add shipping-features orchestrator + capturing-knowledge#9873pa-lem wants to merge 18 commits into
pa-lem wants to merge 18 commits into
Conversation
Re-add the generalized skills from opsmill-skills PR #26 into the loaded .agents/skills/ dir so they can be tested in infrahub before finalizing the plugin PR. Removes the stale, unloaded pre-migration copies under dev/skills/ (feature-flow, capture-knowledge) that PR #9461 already deleted on the migration path. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…ductor
Classification-first (type × size × risk) selects lane + depth. Durable
ship.md manifest in the speckit feature dir is the single source of truth
for resume/reconcile. Layered reliability model (gate everywhere, parallel
divergence at design forks, adversarial verify at correctness claims,
stacked only on risk flags). Discovery table wires in commit/pr/pr-monitor
siblings. Adds phases/{classification,manifest,reliability}.md and a
colleague design brief.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
… skills After rebasing onto develop, the skills it orchestrates now ship in-repo under .agents/. Align references to real names: create-issue→creating-issues, grill-idea→grilling-ideas, pr-monitor→monitoring-pull-requests; bug-* are host commands; adopt the native speckit-review-* suite for the review phase; drop the stale opsmill-software plugin sibling framing and the PR #27 caveat. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…phase Add a post-implement residue-cleanup step (prune leftover debug logs, dead code, orphaned files/imports) before the phase-4 gate, so review sees a tighter diff; re-run the gate test after pruning. Registered in the discovery table with code-simplifier / agent fallback, so it degrades cleanly until the skill resolves in this checkout. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Extend the phase-4 prune step and discovery-table description so the cleanup pass removes redundant/obvious comments alongside debug logs, dead code, and orphaned imports before review sees the diff. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…ge phase Phase 6 now runs /audit-docs after capturing-knowledge — cross-checks the branch's changes against every doc layer and updates drift via /add-docs, so docs stay consistent with what was built. Registered in the discovery table and the classification lane map. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
New phase 10 delegates to speckit-opsmill-retrospect (opsmill:commands/ retrospect.md) — surfaces context-management gaps and routes them to fix-now / open-pr / github-issue / local-only, writing retrospective.md in the feature dir. Distinct from phase 6: knowledge captures domain facts, the retrospective improves the process/tooling surface. Registered in the discovery table and lane map. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Brief now matches SKILL.md + deck: phase 6 is Knowledge & docs (adds /audit-docs → /add-docs), new phase 10 Retrospective (speckit-opsmill- retrospect), prune-residues note at phase 4 incl. comments, the two-cleanups summary, retrospective.md artifact, and an updated status reflecting the first pressure-test pass. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…ge step Phase 6 (knowledge) now runs capturing-knowledge, learning-from-review, the /speckit.opsmill.retrospect retrospective, and the docs audit together, and the standalone Phase 10 is removed (Phase 9 CI-watch becomes terminal). Duplicates the learning-from-review skill from opsmill-skills into .agents/skills/ so it resolves in-repo without the marketplace. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
![cubic-dev-ai[bot]](https://avatars.githubusercontent.com/in/1082092?s=60&v=4){kind=small}
Contributor
There was a problem hiding this comment.
All reported issues were addressed across 3 files (changes from recent commits).
Shadow auto-approve: would not auto-approve because issues were found.
Re-trigger cubic
Contributor
There was a problem hiding this comment.
1 issue found and verified against the latest diff
Confidence score: 4/5
- In
.agents/skills/capturing-knowledge/SKILL.md, the discovery command is scoped todev/anddocs/and omitsknowledge/,guidelines/, andguides/, so repositories using those root-level paths may silently miss relevant documentation and produce incomplete captures; update the command to include all listed bucket paths before merging.
Prompt for AI agents (unresolved issues)
Check if these issues are valid — if so, understand the root cause of each and fix them. If appropriate, use sub-agents to investigate and fix each issue separately.
<file name=".agents/skills/capturing-knowledge/SKILL.md">
<violation number="1" location=".agents/skills/capturing-knowledge/SKILL.md:68">
P3: Discovery command only searches `dev/` and `docs/` but misses the root-level bucket paths listed in the same description (`knowledge/`, `guidelines/`, `guides/`). For a project-agnostic skill, a repo that keeps docs at root level would get a false negative from this command and fall through to "ask the user" unnecessarily.</violation>
</file>
Shadow auto-approve: would not auto-approve because issues were found.
Re-trigger cubic
pa-lem
changed the title
…step Phase 6 now prefers the audit-docs (and add-docs) skills — arriving in infrahub via opsmill-speckit — with the /audit-docs and /add-docs commands as the resolving fallback. Updates the wrap-up step, discovery table, lane map, and the brief to match. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Contributor
There was a problem hiding this comment.
All reported issues were addressed across 2 files (changes from recent commits).
Shadow auto-approve: would not auto-approve because issues were found.
Tip: Review your code locally with the cubic CLI to iterate faster.
Re-trigger cubic
Renumber the pipeline 1–10 across SKILL.md (headers + cross-references),
phases/{classification,manifest,pr-split}.md, and the brief (pipeline table +
artifact/phase references). No behavioural change — labels only.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Contributor
There was a problem hiding this comment.
0 issues found across 5 files (changes from recent commits).
Shadow auto-approve: would not auto-approve. Auto-approval blocked by 7 unresolved issues from previous reviews.
Re-trigger cubic
The learning-from-review skill now ships separately in #9910. shipping-features keeps its delegation references to it (they degrade gracefully when the skill is absent). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
pa-lem
changed the title
…psMill 5-stage framework Collapse the bespoke 10-phase pipeline onto the canonical Intake → Prep → Implement → Delivery → Extract stages, delegating the heavy lifting to the speckit-opsmill-* meta-commands (prep/implement/extract). Keep only the orchestration layer as original: classification (type×size×risk → lane+depth), gate/parallel/verify at the seams, human checkpoints (verify & accept), external single-source-of-truth (ship.md demoted to a pointer index; content lives in Jira/specs/report/docs/PR), Jira propose-on-accept, and a lighter bug lane. Rewrite manifest.md to the index model; align classification/pr-split vocabulary; rewrite the brief. learning-from-review referenced as optional (ships separately in #9910). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Contributor
There was a problem hiding this comment.
All reported issues were addressed across 5 files (changes from recent commits).
Shadow auto-approve: would not auto-approve because issues were found.
Tip: Review your code locally with the cubic CLI to iterate faster.
Re-trigger cubic
pa-lem
changed the title
Contributor
There was a problem hiding this comment.
0 issues found across 2 files (changes from recent commits).
Shadow auto-approve: would not auto-approve. Auto-approval blocked by 5 unresolved issues from previous reviews.
Re-trigger cubic
…0006) Four gaps found while shipping IFC-2867 with this skill: - Intake: verify the base branch contains the ticket's referenced code before cutting (the run first branched off stable, module only on develop). - Manifest: never overwrite a feature.json that points at another feature's local state when the lane runs no speckit steps. - Reliability: give the verify skeptic explicit lenses (correctness, concurrency, guidelines compliance, IO efficiency) even at size S — a single refute prompt missed a lock race, an N+1, and a guideline violation that reviewers then caught. - Delivery: fetch ALL unresolved review threads (bots included) and re-fetch after every push. Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
…en their material exists Retrospect and capturing-knowledge are session-bound: their raw material decays when the session ends and does not depend on review feedback, so they run at the Stage 4 checkpoint (PR open, CI green) on every lane. harvesting-review is artifact-bound and cannot run on a schedule — review feedback arrives after the session — so it anchors to review activity: the Stage 4 review-fix loop harvests each round it just processed, ship.md carries a harvest marker on the delivery row, and resume/reconcile re-opens the step when the PR has unharvested threads (with a final harvest proposed at merge, before Jira Done). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Contributor
There was a problem hiding this comment.
0 issues found across 2 files (changes from recent commits).
Shadow auto-approve: would not auto-approve. Auto-approval blocked by 3 unresolved issues from previous reviews.
Re-trigger cubic
…/create-jira-tickets alignment, arrow spacing Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Contributor
There was a problem hiding this comment.
0 issues found across 2 files (changes from recent commits).
Shadow auto-approve: would auto-approve. Adds agent orchestration skills and removes stale copies; no production code changes, purely tooling/documentation.
Re-trigger cubic
- Stage 4 delivery list renumbered into a single 1→6 sequence (was a duplicated `4` with extraction wedged between the two review steps). - Drop command aliases that resolve nowhere in-repo (`/pr-monitor`, `/create-prd`, `/grill-idea`); use the real skills (`monitoring-pull-requests`, `creating-prd`, `grilling-ideas`). - overview-and-design: move session-bound `retrospect` + `capturing-knowledge` to the Stage 4 cell (matching SKILL), add the actionable-PR-comments step, and add `audit-docs`→`add-docs` for docs. - Stage 5 Extract now references `audit-docs`/`add-docs` for feature docs. - reliability: define "phase" vs "stage" so the gate-every-phase / gate-every-stage wording no longer reads as a contradiction. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Contributor
There was a problem hiding this comment.
1 issue found across 3 files (changes from recent commits).
Confidence score: 4/5
- In
.agents/skills/shipping-features/SKILL.md, the discovery table points contributors to a/reviewcommand that does not exist in.agents/commands/or.agents/skills/, which can mislead agent workflows and cause review steps to fail or be skipped — update the reference to a real command/skill or add the missing/reviewimplementation.
Prompt for AI agents (unresolved issues)
Check if these issues are valid — if so, understand the root cause of each and fix them. If appropriate, use sub-agents to investigate and fix each issue separately.
<file name=".agents/skills/shipping-features/SKILL.md">
<violation number="1" location=".agents/skills/shipping-features/SKILL.md:225">
P2: Custom agent: **Flag AI Slop and Fabricated Changes**
The discovery table references `/review` as an in-repo command, but no such command exists in `.agents/commands/` and no `review` skill exists in `.agents/skills/`. This fabricated reference could mislead users into invoking a non-existent tool. Replace it with an actual in-repo review command or skill, such as `speckit-review-comments` or `/speckit.review.comments`.</violation>
</file>
Shadow auto-approve: would not auto-approve because issues were found.
Tip: Review your code locally with the cubic CLI to iterate faster.
Re-trigger cubic
| | Commit | `commit`, `/git-commit` | `commit-commands:commit` | `git commit` (conventional) | | ||
| | PR | `pr`, `/git-pr` | `commit-commands:commit-push-pr`, `superpowers:finishing-a-development-branch` | `gh pr create` | | ||
| | Post-open CI watch | `monitoring-pull-requests` | — | `gh run watch` / skip | | ||
| | PR review comments | `/review` | `code-review:code-review`, `coderabbit:code-review` | `gh pr review --comment` / `gh api` (drafted findings + prompt blocks) | |
Contributor
There was a problem hiding this comment.
P2: Custom agent: Flag AI Slop and Fabricated Changes
The discovery table references /review as an in-repo command, but no such command exists in .agents/commands/ and no review skill exists in .agents/skills/. This fabricated reference could mislead users into invoking a non-existent tool. Replace it with an actual in-repo review command or skill, such as speckit-review-comments or /speckit.review.comments.
Prompt for AI agents
Check if this issue is valid — if so, understand the root cause and fix it. At .agents/skills/shipping-features/SKILL.md, line 225:
<comment>The discovery table references `/review` as an in-repo command, but no such command exists in `.agents/commands/` and no `review` skill exists in `.agents/skills/`. This fabricated reference could mislead users into invoking a non-existent tool. Replace it with an actual in-repo review command or skill, such as `speckit-review-comments` or `/speckit.review.comments`.</comment>
<file context>
@@ -198,6 +222,7 @@ Use the first available source per capability. Probe; never invent a command the
| Commit | `commit`, `/git-commit` | `commit-commands:commit` | `git commit` (conventional) |
| PR | `pr`, `/git-pr` | `commit-commands:commit-push-pr`, `superpowers:finishing-a-development-branch` | `gh pr create` |
| Post-open CI watch | `monitoring-pull-requests` | — | `gh run watch` / skip |
+| PR review comments | `/review` | `code-review:code-review`, `coderabbit:code-review` | `gh pr review --comment` / `gh api` (drafted findings + prompt blocks) |
| Extract knowledge | `/speckit-opsmill-extract`, `capturing-knowledge`, `audit-docs`/`/audit-docs` → `add-docs` | — | edit `dev/` docs directly |
| Learn from review | `harvesting-review` *(PR #9922, in `stable`)* | — | skip |
</file context>
Suggested change
| | PR review comments | `/review` | `code-review:code-review`, `coderabbit:code-review` | `gh pr review --comment` / `gh api` (drafted findings + prompt blocks) | | |
| | PR review comments | `speckit-review-comments` | `code-review:code-review`, `coderabbit:code-review` | `gh pr review --comment` / `gh api` (drafted findings + prompt blocks) | |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What
Adds a single entry point for delivering one unit of work end to end, plus its knowledge helper, and removes two stale pre-migration skill copies.
shipping-features— a conductor over the OpsMill 5-stage framework: Intake → Prep → Implement → Delivery → Extract. It reimplements nothing — the stages and their tools already exist; this skill adds a thin orchestration layer on top so a developer never has to ask "where am I, which command runs next?"capturing-knowledge— persists non-obvious domain/code facts into the project's knowledge / guideline / guide docs; used standalone and as a delegate during the Extract stage.The orchestration layer (what this skill actually contributes)
The heavy lifting is delegated to the canonical tools —
/speckit-opsmill-prep(specify → plan → critique → tasks),/speckit-opsmill-implement(preflight → clean-context subagents → review-run → report),/speckit-opsmill-extract, plus/pr·monitoring-pull-requests·/pre-ciand the/bug-*commands. On top of that,shipping-featuresadds five things:type × size × risk, which picks the lane and the depth. A bug takes a lighter lane (/bug-analyze → /bug-tdd → /bug-fix); size decides how many steps/checkpoints.M/Lonly) · verify (an adversarial skeptic refutes correctness claims). All three stack on one stage only under a risk flag.S(fast path) but driven granularly onM/Lso the user approves each decision. Never chains a whole stage unattended unless the fast-path is chosen.specs/, build in the implement report, knowledge indev/docs, delivery in the PR).ship.mdis a thin index of pointers, not a store. Resume reads the index and reconciles against the live sources — the source always wins.Relationship to
harvesting-reviewThe review-learning skill shipped as
harvesting-reviewin #9922 (merged intostable) — the earlierlearning-from-reviewdraft (#9910) was closed and superseded.shipping-featuresreferencesharvesting-reviewas an optional Extract-stage step and degrades cleanly when it's absent, so this PR carries no dependency on it.Works without the marketplace
Every tool the pipeline discovers already ships in-repo — sibling skills (
creating-issues,grilling-ideas,creating-prd,speckit-opsmill-*,pr,monitoring-pull-requests,rebase,commit) and commands (/bug-analyze,/bug-tdd,/bug-fix,/speckit-*,/pre-ci,/audit-docs,/add-docs,/git-pr). It runs end to end with no plugins installed; marketplace plugins (superpowers, coderabbit, commit-commands) and built-in agents are optional/fallback layers.Files
.agents/skills/shipping-features/—SKILL.md,docs/overview-and-design.md,phases/(classification, manifest, reliability, pr-split).agents/skills/capturing-knowledge/SKILL.mddev/skills/feature-flow/anddev/skills/capture-knowledge/(superseded byshipping-featuresandcapturing-knowledge)🤖 Generated with Claude Code