Regenerate Probo skill pack with multi-stack guidelines + adversarial reviewer#1139
Draft
aureliensibiril wants to merge 9 commits into
Draft
Regenerate Probo skill pack with multi-stack guidelines + adversarial reviewer#1139aureliensibiril wants to merge 9 commits into
aureliensibiril wants to merge 9 commits into
Conversation
e47b0fc to
d4b5e63
Compare
Signed-off-by: Aurélien Sibiril <81782+aureliensibiril@users.noreply.github.com>
Signed-off-by: Aurélien Sibiril <81782+aureliensibiril@users.noreply.github.com>
Signed-off-by: Aurélien Sibiril <81782+aureliensibiril@users.noreply.github.com>
Signed-off-by: Aurélien Sibiril <81782+aureliensibiril@users.noreply.github.com>
Signed-off-by: Aurélien Sibiril <81782+aureliensibiril@users.noreply.github.com>
Signed-off-by: Aurélien Sibiril <81782+aureliensibiril@users.noreply.github.com>
Signed-off-by: Aurélien Sibiril <81782+aureliensibiril@users.noreply.github.com>
Signed-off-by: Aurélien Sibiril <81782+aureliensibiril@users.noreply.github.com>
…iewer Re-run potion-skill-generator on the full monorepo (61 explored modules + 80 PR review comments + 28 contrib/claude/* docs) to produce a tailored skill pack: - Skills: /potion-ask, /potion-plan, /potion-implement, /potion-review, /potion-learn (mines getprobo/probo PR reviews + drift detection) - Generalist agents: explorer, planner, implementer, reviewer - Stack-specific implementers: go-backend, typescript-frontend (each loads only its stack's guidelines for focused context) - Specialized reviewers (now potion- prefixed): architecture, pattern, security, style, test, duplication, plus opt-in adversarial reviewer wired to OpenAI Codex via the local Codex MCP server - Multi-stack guidelines tree: .claude/guidelines/shared.md + go-backend/ + typescript-frontend/ (3,921 lines, 35 files, including 19 + 10 module-notes) The shared guidelines codify the four-surface API rule (GraphQL <-> MCP <-> CLI <-> n8n) and the 19 review-enforced standards mined from human PR comments. Stack guidelines reference shared.md instead of duplicating cross-cutting rules. The implementer skill enforces the four-surface rule before delegating. AGENTS.md now points to the skill pack as a top-of-file section while keeping contrib/claude/ as the authoritative source-of-truth for subsystem rules. Also: gitignore the local .skill-gen-workspace/ and .claude_context. Signed-off-by: Aurélien Sibiril <81782+aureliensibiril@users.noreply.github.com>
d4b5e63 to
9755b29
Compare
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
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.
Summary
Re-runs
potion-skill-generatoragainst the full monorepo to produce an up-to-date, tailored Claude Code skill pack at.claude/. Replaces the prior generation (skills already prefixedpotion-, but agents and module notes were stale).Inputs digested: 61 modules explored, 80 human PR review comments mined, 28
contrib/claude/*.mddocs absorbed.What changed
Skills (
.claude/skills/):/potion-ask,/potion-plan,/potion-implement,/potion-review,/potion-learn(new — minesgetprobo/proboPR reviews, detects drift betweencontrib/claude/and code, stages guideline updates).Agents (
.claude/agents/): All renamed topotion-prefix.potion-explorer,potion-planner,potion-implementer,potion-reviewerpotion-go-backend-implementer,potion-typescript-frontend-implementer(each loads only its stack's guidelines for focused context).claude/agents/reviewers/): architecture, pattern, security, style, test, duplication, plus opt-in adversarial reviewer wired to OpenAI Codex via the local Codex MCP serverGuidelines (
.claude/guidelines/): 3,921 lines, 35 files in a multi-stack tree.shared.md— cross-cutting: four-surface API rule (GraphQL ↔ MCP ↔ CLI ↔ n8n), GIDs, ISC headers, git workflow, security baseline, 19 PR-review-enforced standards distilled from human reviewer commentsgo-backend/— patterns, conventions, testing, pitfalls + 19 module notes (one per package group)typescript-frontend/— same shape + 10 module notesAGENTS.mdgains a top-of-file section pointing to the skill pack while keepingcontrib/claude/as the authoritative source-of-truth..gitignorenow excludes the local.skill-gen-workspace/and.claude_contextcaches.Why
The four-surface API rule (GraphQL ↔ MCP ↔ CLI ↔ n8n) is the rule most likely to be silently dropped by an AI assistant lacking Probo context. The
potion-implementskill enforces it before delegating; the with/without evaluation showed a vanilla agent literally said "MCP/CLI/n8n not on critical path" while the skill-loaded agent generated concrete diffs across all four surfaces.Other measurable wins from the with/without comparison:
potion-planturns prior-art discovery into aAskUserQuestiondecision (Route A reuse vs Route B new entity) per skill methodologypotion-reviewenforces severity categorization + cites guideline section numbers + caught the license year-range rule that vanilla missedpotion-askfollowed the lazy-loading order exactly and structured failure modes by categoryTest plan
git diff --stat origin/mainshows only.claude/,AGENTS.md, and.gitignorechanges/potion-ask "where is tenant isolation enforced"loadsshared.mdthengo-backend/index.mdthen drills intomodule-notes/coredata.md/potion-plan "add an AuditTrail entity"surfaces the existingAuditLogEntryas Route A/potion-implement "add a pendingActionsCount field"routes topotion-go-backend-implementerand produces concrete diffs across all four surfaces/potion-review "review pkg/agent/tools/search/web_search.go"flags barehttp.Client{}as a security blocker citingshared.md §12/potion-review "second opinion on this PR"triggers the adversarial reviewer (requires Codex CLI + ChatGPT login orOPENAI_API_KEY)Summary by cubic
Regenerated the Probo skill pack under
.claude/with multi-stack guidelines andpotion-‑prefixed skills/agents, plus an opt-in adversarial reviewer via Codex MCP. Planning, implementation, and reviews are now stack‑aware and enforce the four‑surface API rule across GraphQL ↔ MCP ↔ CLI ↔ n8n.New Features
/potion-ask,/potion-plan,/potion-implement,/potion-review,/potion-learn(mines PR reviews, detects guideline drift).potion-explorer,potion-planner,potion-implementer,potion-reviewer), stack implementers (potion-go-backend-implementer,potion-typescript-frontend-implementer), and reviewers (architecture, pattern, security, style, test, duplication,potion-adversarial-reviewervia Codex MCP); allpotion-‑prefixed..claude/guidelines/shared.mdplusgo-backend/andtypescript-frontend/trees with module notes; referencecontrib/claude/.Migration
.claude/,AGENTS.md(points to the pack), and.gitignore(ignores.skill-gen-workspace/and.claude_context).codex loginorOPENAI_API_KEY.Written for commit 9755b29. Summary will update on new commits.