refactor!: ultra-minimal slim pass (drop 14 skills, 21 agents, coding-tutor, copilot/droid targets)#5
Merged
Merged
Conversation
Removed from the plugin; the legacy-cleanup registries now fingerprint and sweep stale installs on upgrade. BREAKING CHANGE: removes the lfg skill. Op: correct Restores: spec:fork-slimness (AGENTS.md Scope & Lineage)
Removed from the plugin; the legacy-cleanup registries now fingerprint and sweep stale installs on upgrade. BREAKING CHANGE: removes the ce-clean-gone-branches skill. Op: correct Restores: spec:fork-slimness (AGENTS.md Scope & Lineage)
Removed from the plugin; the legacy-cleanup registries now fingerprint and sweep stale installs on upgrade. BREAKING CHANGE: removes the ce-report-bug skill. Op: correct Restores: spec:fork-slimness (AGENTS.md Scope & Lineage)
Removed from the plugin; the legacy-cleanup registries now fingerprint and sweep stale installs on upgrade. BREAKING CHANGE: removes the ce-strategy skill. Op: correct Restores: spec:fork-slimness (AGENTS.md Scope & Lineage)
Removed from the plugin; the legacy-cleanup registries now fingerprint and sweep stale installs on upgrade. BREAKING CHANGE: removes the ce-dogfood-beta skill. Op: correct Restores: spec:fork-slimness (AGENTS.md Scope & Lineage)
Removed from the plugin; the legacy-cleanup registries now fingerprint and sweep stale installs on upgrade. BREAKING CHANGE: removes the ce-test-browser skill. Op: correct Restores: spec:fork-slimness (AGENTS.md Scope & Lineage)
Removed from the plugin; the legacy-cleanup registries now fingerprint and sweep stale installs on upgrade. BREAKING CHANGE: removes the ce-demo-reel skill. Op: correct Restores: spec:fork-slimness (AGENTS.md Scope & Lineage)
Removed from the plugin; the legacy-cleanup registries now fingerprint and sweep stale installs on upgrade. BREAKING CHANGE: removes the ce-release-notes skill. Op: correct Restores: spec:fork-slimness (AGENTS.md Scope & Lineage)
…ill and its exclusive agent ce-pr-comment-resolver Removed from the plugin; the legacy-cleanup registries now fingerprint and sweep stale installs on upgrade. BREAKING CHANGE: removes the ce-resolve-pr-feedback skill and its exclusive agent ce-pr-comment-resolver. Op: correct Restores: spec:fork-slimness (AGENTS.md Scope & Lineage)
…t ce-issue-intelligence-analyst Removed from the plugin; the legacy-cleanup registries now fingerprint and sweep stale installs on upgrade. BREAKING CHANGE: removes the ce-ideate skill and its exclusive agent ce-issue-intelligence-analyst. Op: correct Restores: spec:fork-slimness (AGENTS.md Scope & Lineage)
Removed from the plugin; the legacy-cleanup registries now fingerprint and sweep stale installs on upgrade. BREAKING CHANGE: removes the ce-polish skill. Op: correct Restores: spec:fork-slimness (AGENTS.md Scope & Lineage)
…-agent-native-audit skills Removed from the plugin; the legacy-cleanup registries now fingerprint and sweep stale installs on upgrade. BREAKING CHANGE: removes the ce-agent-native-architecture and ce-agent-native-audit skills. Op: correct Restores: spec:fork-slimness (AGENTS.md Scope & Lineage)
Removed from the plugin; the legacy-cleanup registries now fingerprint and sweep stale installs on upgrade. BREAKING CHANGE: removes the ce-optimize skill. Op: correct Restores: spec:fork-slimness (AGENTS.md Scope & Lineage)
Workflow targets plugins/compound-engineering/docs/**, which does not exist; it has never deployed anything from this repo. Op: compress
Brainstorms with no plan lineage and no open work pruned; those with live lineage kept. Git history preserves everything. Op: compress
Plans documenting shipped, merged work pruned; plans tied to open work kept. Git history preserves everything. Op: compress
Cleanup-only verticals removed: converters, types, dedicated tests, detection and cleanup wiring. Shared legacy-cleanup helpers survive for active targets. BREAKING CHANGE: removes copilot and droid converter/cleanup support; users on old copilot/droid installs must sweep stale artifacts manually. Op: correct Restores: spec:fork-slimness (AGENTS.md Scope & Lineage)
Second plugin removed: plugin tree, marketplace entries (all three, parity preserved), release-please package + manifest entries, release metadata/component code, test fixtures. BREAKING CHANGE: removes the coding-tutor plugin from the marketplace and release pipeline. Op: correct Restores: spec:fork-slimness (AGENTS.md Scope & Lineage)
Code-review fan-out consolidated: correctness absorbs reliability, maintainability, and code-simplicity; testing absorbs api-contract and project-standards; niche conditional personas pruned (frontend-races, agent-native, data-migration, deployment-verification). BREAKING CHANGE: removes/renames reviewer agents; stale installs are swept via the legacy-cleanup registries. Op: correct Restores: spec:fork-slimness (AGENTS.md Scope & Lineage)
Document review consolidated from seven single-lens reviewers to three multi-lens passes: feasibility (+design,+security lenses), adversarial-document (+product,+scope lenses), coherence. BREAKING CHANGE: removes/renames reviewer agents; stale installs are swept via the legacy-cleanup registries. Op: correct Restores: spec:fork-slimness (AGENTS.md Scope & Lineage)
Planning research consolidated: repo-research-analyst absorbs git-history and pattern-recognition; architecture-strategist absorbs spec-flow analysis. BREAKING CHANGE: removes/renames reviewer agents; stale installs are swept via the legacy-cleanup registries. Op: correct Restores: spec:fork-slimness (AGENTS.md Scope & Lineage)
Shared specialists consolidated into ce-plan-specialist-reviewer (performance, security, data-integrity lenses) dispatched by ce-compound and ce-plan; figma-design-sync pruned. BREAKING CHANGE: removes/renames reviewer agents; stale installs are swept via the legacy-cleanup registries. Op: correct Restores: spec:fork-slimness (AGENTS.md Scope & Lineage)
Deslop + humanizer pass: ce-work SKILL.md and references compressed with pipeline-review-contract lockstep. Structure, sentinels, triggers, and contracts preserved. Op: compress
Deslop + humanizer pass: ce-code-review SKILL.md and references compressed with review-skill-contract lockstep. Structure, sentinels, triggers, and contracts preserved. Op: compress
Deslop + humanizer pass: ce-plan and ce-brainstorm SKILL.md and references compressed together. Structure, sentinels, triggers, and contracts preserved. Op: compress
Deslop + humanizer pass: workflow-cluster skills compressed. Structure, sentinels, triggers, and contracts preserved. Op: compress
Deslop + humanizer pass: git-cluster skills compressed. Structure, sentinels, triggers, and contracts preserved. Op: compress
Deslop + humanizer pass: support-cluster skills compressed. Structure, sentinels, triggers, and contracts preserved. Op: compress
Deslop + humanizer pass: surviving untouched agent prose compressed. Structure, sentinels, triggers, and contracts preserved. Op: compress
Correct the Plugin Maintenance removal procedure: a removal needs THREE registry edits, not two. The third is the LEGACY_ONLY_SKILL_DESCRIPTIONS / LEGACY_ONLY_AGENT_DESCRIPTIONS fingerprint maps in src/utils/legacy-cleanup.ts. A removed component whose name still ships under a successor is matched against the current bundle, but one with no successor has no live entry to match -- without its description fingerprint the sweep never recognizes the orphan, so it is silently never swept. Fix the matching "two registries" claim in Scope & Lineage too. Scrub now-false statements against the live tree: drop coding-tutor from the intro, Directory Layout, Repo Surfaces, the release-components list (now cli / compound-engineering / marketplace / cursor-marketplace per release-please-config.json), and the commit-scope example label; the repo ships one plugin. Update the stale ce-session-inventory skill reference to its successor ce-sessions. The linked-versions paragraph stays: cli and compound-engineering are still linked. CONCEPTS.md verified clean (generic vocabulary, no removed-component entries). Root CLAUDE.md shim stays. Op: correct Restores: spec:docs-match-code (AGENTS.md Plugin Maintenance)
Update the inventory claim to the slimmed surface: 14 skills, 16 agents, one plugin. Remove the GitHub Copilot and Factory Droid install sections and their cleanup/troubleshooting references; the CLI rejects `cleanup --target copilot|droid` after the Stage 2 converter removal, so those lines were false claims. Keep the five converter install targets (opencode, codex, pi, gemini, kiro) and the still-live qwen/windsurf legacy-cleanup targets, which the CLI continues to honor. Deslop pass: scanned for the fenced patterns; the voiced philosophy prose and operational install steps had no clear violations, so compression came from removing the dead target sections (404 -> 351 lines). Op: compress
Op: correct Restores: spec:no-ghost-references
- reuse (src/utils/legacy-cleanup.ts): replace the locally redefined pathExists helper with the identical exported utility from ./files, removing the duplicate implementation and updating call sites to the shared function. Op: compress
- ce-feasibility-reviewer (design-completeness rating): replace the
template-syntax example ("[Dimension]: [N]/10 ...") with a single
sentence describing the rating logic; preserve the 7/10-or-below
emit gate and "Skip irrelevant dimensions".
- ce-correctness-reviewer (confidence calibration): collapse the
repeated "**Anchor N**:" scaffolding into one shared anchor-level
lead-in plus a per-lens domain-adaptation table; every concrete
calibration example is preserved.
- ce-correctness-reviewer (reliability bullet): drop the redundant
"Check for max attempts, exponential backoff, and jitter." sub-phrase
(the bullet already conveys the concept). Finding metadata listed
ce-plan-specialist-reviewer; located by content, the phrase lived in
ce-correctness-reviewer.
Op: compress
- nested-conditionals (ce-adversarial-document-reviewer.md): flatten the three-level Document-type/Origin conditional structure (and the standalone fourth "Suppress entirely" header) into two top-level protocol outcomes — full-per-Depth (requirements any-origin, or plan+Origin:none) vs the restricted set (plan+Origin:path, suppressing Sections 1 and 4). Per-section restrictions, the suppression hard-rule, section numbering (1/2/2b/3/4/4b/5), and the why-suppress rationale are preserved so behavior and the Confidence-calibration section-number references are unchanged. Op: compress
Run multi-task work as a chain of delegated subagents, dispatching one fresh worker per task and clearing an adversarial audit before the next task starts. Op: extend
Add a verification-gate skill that runs the repo-native verifier, reads back changed files, ghost-checks removed names, and diffs delivery against the original request before any completion claim. Op: extend
|
Important Review skippedToo many files! This PR contains 300 files, which is 150 over the limit of 150. To get a review, narrow the scope: ⚙️ Run configurationConfiguration used: defaults Review profile: CHILL Plan: Pro Run ID: 📒 Files selected for processing (300)
You can disable this status message by setting the Use the checkbox below for a quick retry:
✨ Finishing Touches🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
![gemini-code-assist[bot]](https://avatars.githubusercontent.com/in/956858?s=60&v=4){kind=small}
There was a problem hiding this comment.
Code Review
This pull request streamlines the repository by completely removing the coding-tutor plugin, along with numerous deprecated skills, plans, and documentation files, focusing the codebase on a simplified compound-engineering setup. The review comments correctly identify documentation discrepancies resulting from these removals, specifically pointing out incorrect skill counts in README.md and outdated agent references in docs/skills/ce-code-review.md.
Important
The consumer version of Gemini Code Assist on GitHub is being sunset. Starting June 18, 2026, new organization installations will be blocked, and all code review activity will officially cease on July 17, 2026.
For more details on the timeline and next steps, please review the Help Documentation.
| After installing, run `/ce-setup` in any project. It checks your environment, installs missing tools, and bootstraps project config. | ||
|
|
||
| The `compound-engineering` plugin currently ships 28 skills and 37 agents. See the [full component reference](plugins/compound-engineering/README.md) for the complete inventory. | ||
| The `compound-engineering` plugin ships 14 skills and 16 agents. See the [full component reference](plugins/compound-engineering/README.md) for the complete inventory. |
There was a problem hiding this comment.
There is a discrepancy in the number of skills listed here. The PR description and release:validate check indicate that the plugin now ships with 16 skills (14 survivors + 2 new: ce-subagent-driven and ce-verify-before-complete) and 16 agents. Please update the text to state "16 skills" instead of "14 skills" to ensure accuracy and consistency with the actual component inventory.
| The skill detects you're on a feature branch (no PR yet), resolves the base from `origin/HEAD` (or PR metadata when an open PR exists), and computes the diff. Stage 2 reads commit messages and writes a 2-3 line intent summary. Stage 2b auto-discovers the plan in `docs/plans/` from the branch name and reads its Requirements (R1-R8, U1-U6). | ||
|
|
||
| Stage 3 selects reviewers: the 6 always-on, plus security (auth touched), reliability (background job for token cleanup), data-migration (migration file present), and deployment-verification agent when the migration is risky. Seven or eight reviewers total, dispatched in parallel. | ||
| Stage 3 selects reviewers: the 5 always-on, plus security (auth touched), reliability (background job for token cleanup), data-migration (migration file present), and deployment-verification agent when the migration is risky. Eight or nine reviewers total, dispatched in parallel. |
There was a problem hiding this comment.
This paragraph contains several outdated references and numerical inconsistencies based on the agent consolidation and pruning in this PR: 1. Always-on count: It mentions "the 5 always-on", but the always-on list was consolidated to exactly 3 reviewers (ce-correctness-reviewer, ce-testing-reviewer, and ce-learnings-researcher) as documented in the "Diff-aware persona selection" section above. 2. Pruned/Merged agents: It still references reliability (which has been merged into ce-correctness-reviewer), data-migration (which was pruned outright), and deployment-verification (which was also pruned outright). 3. Total count: It states "Eight or nine reviewers total", but with 3 always-on and 1 conditional (security), the actual total for this scenario is 4 reviewers. Please update this narrative to align with the new consolidated 16-agent architecture.
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 6c94b0af98
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| @@ -29,7 +25,7 @@ import { isSafeManagedPath, pathExists, readJson, sanitizePathName } from "../ut | |||
| import { resolveOpenCodeGlobalRoot } from "../utils/opencode-config" | |||
| import { expandHome, resolveCodexHome, resolveTargetHome } from "../utils/resolve-home" | |||
|
|
|||
| const cleanupTargets = ["codex", "opencode", "pi", "gemini", "kiro", "copilot", "droid", "qwen", "windsurf"] as const | |||
| const cleanupTargets = ["codex", "opencode", "pi", "gemini", "kiro", "qwen", "windsurf"] as const | |||
There was a problem hiding this comment.
When upgrading from a version that supported Bun-installed Copilot or Droid artifacts, removing copilot and droid from this allow-list makes cleanup --target copilot / cleanup --target droid fail in resolveCleanupTargets instead of backing up the stale .github / ~/.copilot / ~/.factory files. Those old files can continue to shadow the native plugin install, so the converter targets can be dropped while still keeping their legacy cleanup paths.
Useful? React with 👍 / 👎.
# Conflicts: # .github/.release-please-manifest.json
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: eadce90b12
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
Comment on lines
89
to
90
| - **API contract breakage** (changed response shapes, removed fields) -- ce-api-contract-reviewer owns these | ||
| - **Migration safety** (missing rollback, data integrity, schema drift) -- ce-data-migration-reviewer owns these |
There was a problem hiding this comment.
After this slim pass, ce-api-contract-reviewer and ce-data-migration-reviewer are no longer shipped, but this reviewer still tells itself not to report those issues because the deleted personas “own” them. When ce-adversarial-reviewer is dispatched on risky diffs, API contract or migration-safety failures can now be suppressed without any remaining specialized reviewer to catch them; redirect these exclusions to the surviving merged personas or remove the suppressions.
Useful? React with 👍 / 👎.
A name match against the legacy allow-list moved user-authored `~/.codex/skills/<name>` dirs to legacy-backup. Gate the move on the same fingerprint ownership check prompts already use: foreign descriptions are preserved, CE-fingerprinted and fingerprint-less (orphan) dirs still sweep. Op: correct Restores: spec:ownership-gated cleanup (prompt-gate parity in cleanupKnownLegacyCodexArtifacts)
The skill ownership gate matches descriptions against current fingerprints only; flat-era installs whose descriptions were later reworded classified as foreign and stranded. Mirror the prompt alias map's contract: every shipped description reword is a recognized CE fingerprint. Op: correct Restores: spec:alias parity with LEGACY_PROMPT_DESCRIPTION_ALIASES contract
The Stage-6 ghost sweep missed three live surfaces: ce-adversarial-reviewer routed don't-flag ownership to four deleted personas (reliability/maintainability -> ce-correctness-reviewer, api-contract -> ce-testing-reviewer per the merge map; the pruned data-migration exclusion is removed so migration cascades fall back to adversarial's emergent-risk charter instead of being suppressed ownerless). Plugin AGENTS.md still instructed contributors about coding-tutor and cited the removed resolve-pr-feedback skill. docs/skills/ce-code-review.md described the pre-merge reviewer roster in two spots. Op: correct Restores: spec:ghost-sweep invariant (cut names live only in registries, CHANGELOGs, and historical records)
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.
What this is
A single ultra-minimal pass over the entire repo: 37 atomic commits, every removal user-approved per item, every commit individually gated on the full test suite +
release:validate.Net: 306 files changed, +3,198 / -36,384 (-33,186 lines).
Inventory before -> after
ce-subagent-driven,ce-verify-before-complete)BREAKING CHANGES
Capability removals, not refactors. Each landed as
refactor(<scope>)!:with aBREAKING CHANGE:footer:Every removed skill/agent is registered in all three legacy-cleanup registries (
STALE_*,LEGACY_ONLY_*_DESCRIPTIONSfingerprints,EXTRA_LEGACY_ARTIFACTS_BY_PLUGIN), so stale flat installs sweep on upgrade. The descriptions registry was the audit-discovered third edit; AGENTS.md now documents all three.Coverage honestly lost
The agent merge reduces actual review coverage; this is the point, not collateral:
What the tests do and do not guarantee
Two-number accounting
Release notes
Removal commits carry
!; release-please will major the linked cli + compound-engineering pair once. This PR title carries!so the signal survives squash-merge.Verification
bun test: 878 pass / 0 fail on tipbun run release:validate: in sync (16 agents, 16 skills)