chore: initialize marketplace scaffold#1
Merged
Conversation
Fresh public marketplace skeleton: marketplace.json catalog, README, CLAUDE.md fresh-docs mandate, migration playbook, plugins/ dir. Co-Authored-By: Claude <noreply@anthropic.com>
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 5a7469e640
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
+54
to
+55
| - **Namespacing.** An in-repo `/foo` becomes `/melodic-software:foo` (plugin-namespaced). Internal | ||
| cross-references to the bare name break — update them. |
There was a problem hiding this comment.
When this playbook is used to migrate a skill whose plugin name is not exactly melodic-software, it tells maintainers to rewrite slash-command references to the marketplace name. I checked the current Claude Code plugin marketplace docs, and installed plugin skills are namespaced by the plugin name (for example /quality-review-plugin:quality-review after installing quality-review-plugin@my-plugins), so references like /melodic-software:foo will not resolve for plugins published under this marketplace. Please make the example use the actual plugin name, e.g. /<plugin-name>:foo.
Useful? React with 👍 / 👎.
This was referenced Jun 28, 2026
This was referenced Jul 4, 2026
This was referenced Jul 11, 2026
Merged
This was referenced Jul 11, 2026
feat: add review-toolkit plugin (6 reviewer agents + quality-gate and code-review-fanout skills)
#71
Merged
This was referenced Jul 12, 2026
kyle-sexton
added a commit
that referenced
this pull request
Jul 12, 2026
…r self-contained Address three more Codex P2 findings: - teach topic case (#1): topic is start-or-resume, so a fresh-workspace expectation flakes on a machine with prior data. Assert the start-or-resume contract — mission interview for a new workspace, resume for an existing one. - teach primer case (#3): the primer intakes the user's starting point with one question before building the ladder, so a single-turn run may correctly open with the intake. Make the expectation intake-first. - diagnose cleanup case (#6): a broad grep -r "[DEBUG-" already matches the diagnose docs and this fixture in the checkout. Scope the assertion to the specific [DEBUG-a4f2] tag across the changed source.
This was referenced Jul 12, 2026
kyle-sexton
added a commit
that referenced
this pull request
Jul 13, 2026
… hardening (#138) Applies the resolved `planning/setup` hardening to `knowledge/setup`'s `library_dir` seam (same single-directory `userConfig` shape, same latent issues). ## Changes **`plugins/knowledge/skills/setup/SKILL.md`** - **Precedence** — step 1 now resolves the full documented order (Managed > `--settings` > Local > Project > User), warns when higher layers are unreadable instead of declaring a lower value authoritative, and surfaces layer shadowing before the project write. - **Portability** — step 5 guards the tracked project write: never propagate a machine-specific personal-layer value into team settings; route personal-only overrides to `.claude/settings.local.json` (portability required only for the shared project write). - **Declared-convention alignment** — inspect the repo's declared working-notes/artifacts convention even when a value is set; surface divergence and offer to reconcile. **`plugins/codebase-audit/skills/setup/SKILL.md`** (partial match — writes a tracked `.md`, not settings) - Applied only the precedence/shadowing subset: step 1 reports the *effective additively-merged* config across user-global → team → local layers (and warns on unreadable layers), instead of summarizing the team file alone. Settings-write portability specifics do not apply. ## Deviations (own-flagged) 1. **Fix #3 premise softened for knowledge.** The issue's fix #3 parenthetical assumed knowledge "honors that convention over the config at write time" (true for planning). Verified via `skills/youtube/extraction/lib/work-root.js` and the youtube pipeline that **knowledge resolves `library_dir` directly at write time — no skill overrides it with a repo convention at runtime.** Documenting that behavior would be false, so I softened the imported write-time-override claim: `library_dir` is runtime-authoritative, and setup's job is to keep it *aligned* with any declared convention. Fixes #1 and #2 are unaffected. 2. **`codebase-audit` version bump not in issue scope.** Patch-bumped `0.1.0 → 0.1.1` because its SKILL behavior changed; the issue only specified the `knowledge` minor bump. ## Verification - `knowledge` `plugin.json` minor bump `0.2.0 → 0.3.0`; `codebase-audit` `0.1.0 → 0.1.1`. - `claude plugin validate` clean on both plugins. - markdownlint clean on both edited files. Refs melodic-software/medley#1463 <!-- CURSOR_SUMMARY --> --- > [!NOTE] > **Low Risk** > Changes are limited to plugin skill documentation and version metadata; no application runtime or security-sensitive code paths are modified. > > **Overview** > **Hardens setup skills** for the `knowledge` and `codebase-audit` plugins so interactive setup reflects layered config reality instead of treating a single team file as authoritative. > > For **`knowledge/setup`**, step 1 now follows the full Claude Code precedence chain (Managed → `--settings` → Local → Project → User), warns when higher layers are unreadable, and calls out when a project write would stay shadowed. A new step reconciles `library_dir` with repo-declared working-notes conventions even when a value is already set. Persistence (step 5) blocks copying machine-specific personal paths into tracked project settings, routes personal-only overrides to `.claude/settings.local.json`, and the output section documents honest pipeline behavior (`book-distill` ignores `library_dir`; YouTube does not yet honor it). **`knowledge`** bumps **0.2.0 → 0.3.0**. > > For **`codebase-audit/setup`**, step 1 is expanded to summarize the **effective merged** config across user-global → team → local layers (including dimension opt-outs via empty source lists), explain that only the team file is written, and warn when overlays block re-enabling dimensions. **`codebase-audit`** patch-bumps **0.1.0 → 0.1.1**. > > <sup>Reviewed by [Cursor Bugbot](https://cursor.com/bugbot) for commit f13139a. Bugbot is set up for automated code reviews on this repo. Configure [here](https://www.cursor.com/dashboard/bugbot).</sup> <!-- /CURSOR_SUMMARY -->
kyle-sexton
added a commit
that referenced
this pull request
Jul 14, 2026
…ntion (#175) ## What Adds a durable guardrail against fragile Claude Code permission grants, in two parts: - **(a) Enforceable check** — a new `permission-hygiene` skill in the `claude-config-audit` plugin, matching the plugin's established idiom (deterministic detector script + `.test.sh` contract test + `reference/criteria.md` + `evals/evals.json`). A `permission-rule-check.sh` detector scans skill/command/agent frontmatter `allowed-tools` and `settings.json` / `settings.local.json` `permissions.allow`, and flags three anti-patterns. - **(b) Convention doc** — `docs/conventions/permission-rule-hygiene/` (README + CHANGELOG), matching the existing `hook-telemetry` / `ecosystem-commands` convention-directory house style, stating the principle, the three anti-patterns, the correct pattern, and the operator-setup boundary — each with official-doc citations. The skill's criteria link to it (reference, don't restate). ## The three anti-patterns → the correct pattern All three make a grant silently do nothing. Verified against current official docs (URLs below): 1. **Interpreter-wildcard / blanket allow rules are dropped in auto mode.** Per permission-modes: "On entering auto mode, broad allow rules that grant arbitrary code execution are dropped: Blanket `Bash(*)` or `PowerShell(*)`; Wildcarded interpreters like `Bash(python*)`; Package-manager run commands; `Agent` allow rules. Narrow rules like `Bash(npm test)` carry over." So a frontmatter grant such as `Bash(python "*helper.py":*)` grants nothing under auto mode. Empirically, a guarded merge helper granted this way was denied even when invoked bare. 2. **Hardcoded absolute machine/user paths.** Bash rules match the command string literally — no `~`/`$HOME`/env expansion — so `Bash(/c/Users/<name>/.../x.sh:*)` breaks on other machines/usernames and leaks a username into source control. 3. **Assuming a skill or plugin can self-grant.** Skill `allowed-tools` is skill-scoped and (per #1) ineffective for auto-mode-gated actions; a plugin `settings.json` supports only the `agent` and `subagentStatusLine` keys, so a `permissions` block there is inert; and an agent editing its own settings to self-grant is blocked (`.claude/` is a protected path; `defaultMode: auto` is ignored from project/local settings so a repo can't grant itself auto mode). **Correct pattern:** expose the guarded helper as a stable bare command on the Bash tool PATH (pre-plugin: a PATH shim in a dir already on PATH; post-migration: the plugin's `bin/`), allow the bare name narrowly (`Bash(babysit_merge.sh:*)` — carries over into auto mode like `Bash(npm test)`, machine -independent, identical before/after migration), and have the **operator** add that bare-name rule once to user-global `~/.claude/settings.json`. ## The check (detector → criteria → evals) `permission-rule-check.sh` (advisory, exits 0; `--count` for a count; requires `jq`) flags: - **P1** interpreter-wildcard / blanket rules (`Bash(*)`, `Bash(python*)`, `Bash(bash <path>*)`, `Bash(sh -c*)`, package-manager runners, `Bash(*.py:*)`). Narrow rules (`Bash(npm test)`, `Bash(babysit_merge.sh:*)`) are NOT flagged — a negative fixture proves this. - **P2** hardcoded machine/user paths (`/c/Users/…`, `/home/…`, `/Users/…`, `C:\Users\…`); `${CLAUDE_PROJECT_DIR}`/`~/` forms are exempt. - **P3** a plugin `settings.json` that declares an inert `permissions` block. `settings.local.json` is parsed for its `permissions.allow` array only — never read or echoed wholesale (matching the sibling `settings-audit` secret-handling posture). Scope vs `settings-audit` is explicit: this skill owns grant portability + auto-mode durability + who adds the operative rule; file correctness (baseline deny/ask, deprecated `:*`, drift) routes to `settings-audit`. Detector and test fixtures assemble machine-path strings from fragments (and the docs use `<name>` placeholders) so no contiguous machine-path literal sits in a committed file — the repo's own machine-specific-path CI lane stays clean without a CI exclude. ## Empirical evidence the guardrail is needed Running the new detector against this marketplace surfaced **six pre-existing interpreter/runner-led frontmatter grants** across unrelated plugins (shapes like `Bash(bash <script>:*)`, `Bash(bash <dir>/*)`, and `Bash(npx:*)`) — none of them the portable bare-name pattern, and the broad forms among them (a globbed script target, a package runner) are exactly what auto mode drops. (For a grant that invokes one fixed script through an interpreter, the detector flags the same authoring anti-pattern without asserting the drop; the bare-name fix is identical.) Those live in other contributors' plugins and are left for a separate follow-up (out of this PR's lane). ## Tests - `permission-rule-check.test.sh`: 27 checks pass (positive + negative fixtures for all three checks, `--count`, `--help`, jq-missing → exit 2, and the narrow-rule-not-flagged negative case). - Full `scripts/run-plugin-tests.sh` suite green; `shellcheck -x` and `shfmt -ci -i 2` clean; markdownlint clean; scripts carry the executable bit. ## Sources (verified this session) - Auto-mode drop / decision order — https://code.claude.com/docs/en/permission-modes - `classifyAllShell`, narrow-rule carryover — https://code.claude.com/docs/en/auto-mode-config - Literal matching, wildcard / `:*` semantics — https://code.claude.com/docs/en/permissions - `allowed-tools` scope, `${CLAUDE_PROJECT_DIR}` substitution — https://code.claude.com/docs/en/skills - Plugin `bin/` on PATH + `agent`/`subagentStatusLine`-only `settings.json` — https://code.claude.com/docs/en/plugins-reference 🤖 Generated with [Claude Code](https://claude.com/claude-code) https://claude.ai/code/session_013gKgRoW8zkFadbE2hE4r4b <!-- CURSOR_SUMMARY --> --- > [!NOTE] > **Low Risk** > Changes are documentation plus read-only local scanning scripts; they do not alter Claude Code behavior or auto-modify user settings. > > **Overview** > Adds a **permission-rule-hygiene** marketplace convention (`docs/conventions/permission-rule-hygiene/`) and a fourth **`claude-config-audit`** skill, **`permission-hygiene`**, that audits whether `allowed-tools` and `permissions.allow` grants are portable and still effective in auto mode. > > The skill runs a deterministic **`permission-rule-check.sh`** detector (plus **`permission-rule-check.test.sh`**, **`reference/criteria.md`**, and **`evals/evals.json`**) that flags **P1** interpreter/blanket/`Agent` grants auto mode drops, **P2** hardcoded user-home paths, and **P3** inert `permissions` blocks in plugin `settings.json`. It is **report-only** (remediation is operator manual: bare command on PATH + user-global allow rule), with scope filters and explicit routing of file-correctness work to **`settings-audit`**. > > Plugin manifest and docs move **`claude-config-audit`** to **0.3.0** and update the root catalog to describe four audit skills. > > <sup>Reviewed by [Cursor Bugbot](https://cursor.com/bugbot) for commit 867e5f2. Bugbot is set up for automated code reviews on this repo. Configure [here](https://www.cursor.com/dashboard/bugbot).</sup> <!-- /CURSOR_SUMMARY --> --------- Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This was referenced Jul 21, 2026
kyle-sexton
added a commit
that referenced
this pull request
Jul 21, 2026
…th routing Eval #1 (create-routing-enterworktree-last) still asserted the old EnterWorktree(name: ...) final action. The create action now routes through the shared worktree-create.sh helper and calls EnterWorktree(path: ...) on the printed path. Update the expected_output prose and add an expectation locking the helper-routing / path: invariant so the eval no longer rewards the in-repo name: path this change removes.
This was referenced Jul 22, 2026
This was referenced Jul 22, 2026
This was referenced Jul 22, 2026
This was referenced Jul 23, 2026
kyle-sexton
added a commit
that referenced
this pull request
Jul 23, 2026
…tion Codex P2 on PR #1160: the interim .claude/architecture/ default bypassed the marketplace topic-docs convention — it ignored memory_dir overrides and created a second, unignored generated surface under .claude, which the convention reserves for observability only. The deepening candidate artifact is a cross-stage handoff (a planning step consumes its agreed-shape entry), so it now resolves through the convention: memory tier, <memory_dir>/<topic-slug>/deepening-candidates-<timestamp>.md, default .work/<topic-slug>/, self-ignored — which also structurally answers the security review's gitignore concern (scan output cannot leak into git history) and the review's directory-creation note (slice creation + self-ignore guard are convention-owned, now cited). - New plugins/architecture/reference/topic-docs.md delta binding - deepening.md Phase 2 resolves via the binding - Eval #1 expectation matches - README Persistence section rewritten - topic-docs convention 2.1.0: architecture row in Implementers table Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
kyle-sexton
added a commit
that referenced
this pull request
Jul 23, 2026
…t path (F1) (#1160) Closes #1156 ## Summary The deepening lens's durable candidate artifact defaulted to `${CLAUDE_PLUGIN_DATA}/deepening-candidates-<timestamp>.md` — broken twice over, verified against current official docs (fetched 2026-07-23): 1. **The token does not substitute in skill markdown content.** The skills reference's "Available string substitutions" set is `${CLAUDE_SESSION_ID}`, `${CLAUDE_EFFORT}`, `${CLAUDE_SKILL_DIR}`, `${CLAUDE_PROJECT_DIR}` + named args. `${CLAUDE_PLUGIN_DATA}` is a path substitution for hook/monitor commands and MCP/LSP server configs only (plugins-reference "Environment variables"). A consumer following the default literally writes to an unexpanded literal `${CLAUDE_PLUGIN_DATA}/…` directory — observed in the SW2030 consumer audit that produced this finding. 2. **Wrong scope even if it resolved.** `${CLAUDE_PLUGIN_DATA}` → `~/.claude/plugins/data/{id}/` — plugin-global, so per-codebase candidates would collide across projects. Changes (architecture 0.3.5 → 0.3.6): - `actions/deepening.md` Phase 2 default → `${CLAUDE_PROJECT_DIR}/.claude/architecture/deepening-candidates-<timestamp>.md` (per-project). The honor-project-convention override branch is unchanged — it worked correctly in the audit run and masked the broken default. - `evals/evals.json` eval #1 expectation updated to match. - `README.md` Persistence section updated to name the real default path. - CHANGELOG + version bump. ## Test plan - `scripts/check-changelog-parity.sh --check-bump origin/main` — pass (0.3.6 has a CHANGELOG entry). - `scripts/check-changed-skills.sh origin/main` — PASS, 0 errors (1 pre-existing Gotchas warning, scoped to #1158). - `grep -rn CLAUDE_PLUGIN_DATA plugins/architecture/` — no remaining references. ## Related Refs #1157, #1158 — follow-up findings from the same consumer audit (handoff-inbox item `20260723-092605-architecture-improve-deepening-audit`), serialized behind this PR. 🤖 Generated with [Claude Code](https://claude.com/claude-code) --------- Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
kyle-sexton
added a commit
that referenced
this pull request
Jul 23, 2026
…ld (PR #1162 review) Two Important review findings on the scan-briefing change: - scan-briefing.md intro declared "four parts" but the template has five sections; the omitted one is §4 badge-acceptance heuristics — exactly the scan-time calibration this change exists to inject (F4). A caller following the stated count could send a briefing without it. Now "five parts", heuristics enumerated. (claude review #1 + Codex P2) - Phase 1.5 gated on candidates that "will carry a Strong badge" and their shallow-signal, but at Phase 1.5 only the scan-time confidence field exists (strong|worth-exploring|speculative) — the recommendation badge is assigned in Phase 2. Gate now keys on `confidence: strong`, and Phase 2 states the confidence->badge mapping explicitly, resolving the field-name and case-style mismatch. (claude review #2) Also sharpened the durable artifact's deepening vs shallow-signal field descriptions with contrasting examples so they aren't filled redundantly (claude review nit). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
3 tasks
This was referenced Jul 24, 2026
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.
Initialize the fresh public marketplace.
Scaffold only; plugins migrated one at a time.