apache/magpie

For a batch of findings from a non-security audit tool (<audit-tool> — ruff / flake8 / mypy / pylint / CodeQL / Apache Verum / Apache Caer / equivalent; full list in the body) against <upstream>, draft the smallest fix for each finding. Re-runs the tool after each batch to confirm the findings are cleared. Produces a commit and a hand-back artefact; never opens a PR on autopilot or merges.

Read-only audit of GitHub Actions workflow runner compatibility for one repository, an explicit repository set, one Apache project with multiple repositories, or the full Apache GitHub org. Finds obsolete GitHub-hosted runner labels and macOS runner/tool architecture mismatches. Produces TSV evidence files; never edits workflows, opens PRs, or posts comments.

Post-vote committer and PMC onboarding for Apache projects. Walks the nominator through every step from ICLA check to welcome announcement for both incubating podlings and graduated top-level projects.

Read-only GitHub activity card for a named contributor on <upstream>. Fetches PR authorship, code-review activity, issues, and PR/issue comments over a configurable window. Limited to GitHub-visible activity — the body documents the off-GitHub tracks the nominator must supply separately. No readiness verdict is produced; use contributor-nomination for a full nomination brief.

Read-only nomination brief for a named GitHub contributor on <upstream>. Aggregates GitHub activity across all contribution tracks plus maintainer-supplied off-GitHub signal, and flags vendor-neutrality context — the evidence a PMC needs to open a committer or PMC nomination thread.

Measures contributor-sentiment signals on <upstream> over a configurable window: thread tone (first-response classification), time-to-first-reply (median hours), first-PR retention (second-PR rate), and reviewer load (Gini coefficient). Compares each signal against a pre-adoption baseline and produces a structured gate report used to decide whether a skill family is ready to advance from experimental to stable.

Read-only readiness tracker that maps a contributor's GitHub activity against the adopter's PMC-declared committer or PMC thresholds and surfaces a traffic-light brief (Not yet / Approaching / Ready to nominate) plus the specific evidence gaps that remain.

Read-only dependency vulnerability audit for one repository or a local checkout. Detects the project's dependency manager(s), runs the appropriate audit tool, surfaces patchable findings grouped by severity, and proposes upgrades for maintainer review. Never modifies manifests or lock files and never opens update PRs.

Read-only license audit of a project's direct and transitive dependency tree. Detects the dependency manager(s), resolves each dependency's declared license from ecosystem metadata, classifies each against a configured policy (ASF three-category A/B/X model or a custom allowlist), and surfaces incompatible, forbidden, and unknown-license dependencies for maintainer review. Never modifies manifests or lock files.

Read-only flaky-test detection from GitHub Actions CI run history for one repository. Parses workflow run outcomes over a configurable window, computes per-job failure rates, and distinguishes intermittent failures (flaky) from consistent failures (deterministically broken). Produces a prioritised triage list without modifying any test code, workflow file, or tracker state.

Draft a single net-new *good first issue* on the configured <upstream> repo from one supplied candidate such as a known gap or a small maintainer-named task. The skill first runs a suitability gate to confirm the candidate is small and newcomer-safe. If it passes the skill drafts one issue. The draft carries scope, code pointers, contributing-doc links, acceptance criteria, and an effort estimate. A readiness checklist gates the draft before it is shown. Nothing is filed via gh until the maintainer explicitly confirms. The skill never curates or relabels the existing backlog.

Sweep the open <issue-tracker> backlog for existing issues that could be labelled as good first issues. Classifies each candidate as READY (propose the GFI label), NEAR-MISS (surface edits to make it GFI-ready), or SKIP using the G1–G7 suitability rubric. Applies labels only after explicit maintainer confirmation; never edits issue bodies without the maintainer's direction.

Read-only maintainer dashboard for the open general-issue backlog of <issue-tracker>. Surfaces a health rating, prioritised recommendations, age and staleness breakdowns, area pressure ranking, and a triage-funnel summary. Output is HTML by default; markdown fallback available.

Merge two open <issue-tracker> issues that describe the same root cause, preserving both reporters' context. Proposes a closing comment on the duplicate and a cross-reference comment on the kept issue. Waits for maintainer confirmation before posting anything or closing anything.

For a single triaged <issue-tracker> issue confirmed as a bug or feature, draft a fix against <upstream> on <default-branch>. Produces the failing test, the smallest production change, the targeted+module test runs, and the commit. The PR is NOT opened on autopilot; the human committer reviews, signs, and pushes. Hand-back artefact summarises branch, commits, test results, and scope.

Sweep a configured pool of resolved or end-of-life <issue-tracker> issues and re-assess each against the current <default-branch>. Per-issue: invoke issue-reproducer to extract and run the reporter's code, classify the runtime outcome, attach a nature analysis, compose a verdict.json. Hand-back-on-completion contract: no comments posted, no transitions, no closures.

Read-only dashboard over a directory of verdict.json files produced by issue-reassess campaigns. Surfaces a health rating, classification distribution, partial-fix surfaces, oldest-unresolved buckets, and per-component breakdowns. Output is HTML by default; markdown fallback available. Read-only on tracker state; consumes campaign artefacts.

For a single <issue-tracker> issue identifying a code-level bug, extract the reporter's example code from the issue body, adapt it to run on the current <default-branch>, execute via <runtime>, and compose a verdict.json describing the observed behaviour vs the expected failure. Read-only on the tracker — produces evidence, never posts. Invoked by issue-triage and issue-reassess; can also be run standalone.

Sweep open <issue-tracker> issues for inactivity past a configurable threshold and propose either a closure (when the issue has been unresponsive long enough to presume abandonment) or an update request (nudge the reporter to confirm the issue is still relevant). Waits for maintainer confirmation before posting any comment or closing anything.

For each open <issue-tracker> issue in the configured candidate pool, read the issue body and comments and classify the candidate disposition. On user confirmation, posts a triage-proposal comment that invites the project team to react. Read-only on tracker state — no workflow transitions, closures, or label changes. Six classes in the body.

Read-only license compliance audit for one repository or a local checkout. Checks that a LICENSE file exists, that a NOTICE file is present and complete when required by the declared license, and that source files carry SPDX-License-Identifier headers consistent with the project's declared license. Produces a grouped compliance report and proposes remedies for maintainer review. Never modifies any file.

Print a human-readable index of every skill in this repository, grouped by family prefix (pr-management, security, setup, …) with each skill's name and the first sentence of its description. The listing is generated on every run from the live .claude/skills/*/SKILL.md files, so it never goes stale when skills are added, removed, or rewritten.

Draft a first-contact orientation comment for a first-time contributor on a newly opened issue or PR on the configured <upstream> repo. Detects first-time authorship via the GitHub author_association field and drafts a welcome with contributing-guide link, community-norm pointers, and expected next steps. Waits for explicit maintainer confirmation before posting. Does not post for repeat contributors.

Given an open good-first-issue on the configured <upstream> repo, explain it in beginner terms and sketch a concrete approach: which files to read first, what "done" looks like, and where to ask follow-up questions — without writing any code or fix. First runs an issue assessment to confirm the issue is open, non-security, and scope-clear. Then drafts the explanation for maintainer review. Read-only; nothing is posted without explicit maintainer confirmation.

Answer a newcomer's "how do I contribute here" question by grounding the response in CONTRIBUTING.md and the project's own docs. Classifies the question (setup / workflow / first-issue / out-of-scope), retrieves the relevant guide excerpt, and drafts a concise answer in the Agentic Mentoring teaching register. Routes out-of-scope questions (design, security, deprecation, architectural taste) to a human maintainer. Read-only; produces a draft answer text — no files are written and no comments are posted without the maintainer's explicit action.

Optimize an existing framework skill (or sweep a set of them) by applying the restructuring patterns proven on the security-skill suite: split an oversized SKILL.md into linked sibling docs, lift concrete/project-specific values out of the body into <project-config> placeholders, replace in-agent-context body reads with out-of-context tool calls, batch per-item fetches into a single upfront pass, and add a deterministic pre-flight no-op classifier ahead of LLM passes. Every change is a behavior- preserving proposal the maintainer signs off on; the skill validator must stay green before and after. The refactoring sibling of write-skill (which authors net-new skills).

Fan a local diff through three independent, axis-focused review passes (correctness, security, conventions), then merge the findings into a single structured report. Each pass is isolated so findings from one axis cannot suppress or bias the others. The merged report uses the same format as pairing-self-review so the developer gets a consistent signal regardless of which Agentic Pairing skill they invoke.

Run a structured pre-flight self-review on local changes before opening a PR. Reads the diff against a configurable base (default: the merge base of HEAD and the upstream default branch), checks correctness, security, and project conventions, and returns a structured report to the developer. No state changes, no PR, no external writes — the report is the output.

Walk a maintainer through deep, sequential code review of open pull requests on the configured <upstream> repo. Defaults to the **"my reviews"** queue (the union of five maintainer signals — see the Inputs table); selectors can narrow to a single PR, an area label, or a collaborator subset. Drafts an approve / request-changes / comment review per PR and posts on the maintainer's confirmation.

Draft a teaching-register comment on a single GitHub issue or PR thread on the configured <upstream> repo, aimed at a contributor who is missing repo context the maintainer would otherwise have to spell out. The skill reads the thread, decides whether a mentoring intervention is warranted, drafts one comment per the project's tone guide and convention pointers, and waits for explicit maintainer confirmation before posting via gh. Escalates to the configured maintainer team on the four hand-off triggers.

Identify trivial, low-risk pull requests in the ready for maintainer review queue of <upstream> that pass every quality gate and touch only supplementary areas (docs, changelog, translations, tests) — the "express lane" a maintainer can review and merge in seconds. Surfaces and ranks candidates with per-PR diff summaries, an all-gates-green attestation, and the exact merge command. On the maintainer's explicit per-PR confirmation it can submit an APPROVE review (the maintainer's own review of the trivial diff — useful when the PR has no approvals yet and branch protection needs one), exactly as pr-management-code-review does. It never merges itself — automated merge is the framework's deliberately-deferred Agentic Autonomous mode; the maintainer runs the printed merge command in their own session.

Read-only maintainer dashboard for the open-PR backlog of <upstream>. Surfaces a health rating, prioritised action recommendations, weekly closure velocity trends, area pressure ranking, and a triage-funnel breakdown — with the underlying area-grouped tables as a collapsible details section.

Sweep open pull requests on the configured <upstream> repo, classify each one against the project's quality criteria, propose a disposition, and — on the maintainer's confirmation — carry out the action via gh. Disposition options per PR: draft / comment / close / rebase / CI-rerun / workflow-approve / ping-stale-reviewer / request author confirmation of readiness / mark ready for maintainer review / promote bot-authored draft. Does **not** perform code review — that lives in pr-management-code-review.

Sweep open pull requests on the configured <upstream> repo for inactivity past a configurable threshold and propose either a conversion to draft (when the PR is open but has gone quiet) or a closure (when the PR has been abandoned long enough to presume the author has moved on). Waits for maintainer confirmation before converting or closing anything.

Run a newcomer-focused pre-flight checklist on a local branch before opening a pull request. Checks CONTRIBUTING conventions, SPDX headers on new files, commit-message shape (including the Generated-by: trailer for AI-assisted work), and the placeholder convention — then returns a structured checklist report. Read-only; no state changes, no PR, no external writes.

Draft the [ANNOUNCE] email body and open (not merge) the site-bump PR for a promoted release of <upstream>. Reads release metadata from the planning issue and <project-config>/release-management-config.md; produces a ready-to-copy [ANNOUNCE] subject + body and proposes the site-bump PR. Never sends mail and never merges the PR without explicit RM confirmation.

Scan the release distribution area (dist/release/<project>/ when release_dist_backend = svnpubsub, or the configured distribution location), identify releases past the project's retention rule, and propose the backend-shaped command set to move them to the archive area. Read-only on the distribution surface; the RM executes every archival command as themselves.

Assemble a per-release audit record from lifecycle artefacts (planning issue, vote thread, artefact list, promote revision, and announcement URL) and propose a PR appending it to the project's audit log. Read-only on every release surface; the only write is a PR the RM reviews and a committer merges.

Draft the diff that adds the Release Manager's public key to the project's KEYS file (<keys-file-url>), emit a paste-ready svn (or backend-equivalent) command sequence, remind the RM to upload to the configured keyserver, and validate the key meets the ASF strength floor. Never commits, never holds or reads the private key. Runs during release preparation, before RC signing begins.

Draft release preparation artefacts for <upstream>: the planning issue, the version-bump and changelog prep PR, or the post-release development-version bump PR. Reads release metadata from <project-config>/release-trains.md and <project-config>/release-management-config.md. Every output is a draft confirmed by the Release Manager before filing; the agent never marks a PR ready, never merges, and never closes any artefact.

Emit the backend-shaped promotion command set for a release that has passed its vote. Reads the planning issue (must carry vote-passed), constructs the staging → release move for the configured distribution backend, checks PMC membership of the Release Manager, and proposes the promoted label. Never runs the promotion command itself and never publishes the release.

Emit the paste-ready command sequence to tag an RC, build artefacts, sign each artefact, generate checksums, and stage them to the adopter's distribution backend. Covers Steps 4–5 of the release-management lifecycle. Never runs any command locally — all sequences are emitted for the Release Manager to execute on their own machine with their own key and ASF credentials.

Read-only pre-flight verification of a staged release candidate (RC) for <upstream>. Checks artefact integrity (GPG signatures and checksums), Apache RAT licence headers, NOTICE/LICENSE completeness, prohibited-binary absence (including .pyc / __pycache__), source-tree integrity (no dangling symlinks or broken internal references), and version-string consistency. Emits a structured PASS / PASS-WITH-WARNINGS / FAIL report. Makes no state change; a --post-to <planning-issue> flag proposes a comment for explicit RM confirmation before any posting.

Draft the [VOTE] email body and planning-issue comment for an RC of <upstream>. Reads RC metadata from the planning issue and <project-config>/release-management-config.md; produces a ready-to-copy [VOTE] subject + body and a proposed planning-issue comment. Never sends mail and never posts without explicit RM confirmation.

After the approval window closes, fetch the approval signal for an RC of <upstream>, classify each reply as +1 / 0 / -1 and binding or non-binding against the configured roster, produce the tally summary, and draft the [RESULT] [VOTE] email. Never sends mail and never applies a label without explicit RM confirmation.

Help an adopter or framework developer file a clean, redacted GitHub issue against the Apache Magpie framework repo when a skill, tool, or doc misbehaves. It gathers the problem from the user — never from the raw session transcript — then runs a mandatory public-disclosure scrub before rendering the report into the framework's bug_report / change_proposal issue template, checking for duplicates, and filing via gh issue create --web only on explicit confirmation. The scrub is the point: the destination is a public repo, so the skill strips any private tracker, embargoed-CVE, private-list, or cross-project content the report would otherwise leak.

Given an open issue or PR, scores the project's configured reviewer roster across three signals — touched-area eligibility, git-history familiarity with the changed paths, and current open-review load — and proposes a primary reviewer (plus an optional backup). Read-only and propose-then-confirm: nothing is assigned, labelled, or requested without the maintainer's explicit confirmation. An unresolved roster produces an explicit NO ELIGIBLE REVIEWER signal, never a fabricated handle.

Walk a security team member through allocating a CVE for an <tracker> tracking issue (governance-gated per governance.cve_allocation_gate). Prints the configured <cve-tool> allocation URL, waits for the allocated CVE ID, then updates the tracker in place. Tracker updates: CVE tool link field, cve allocated label, status-change comment, CVE JSON. Chains into security-issue-sync afterwards to reconcile the rest of the tracker.

Merge two <tracker> tracking issues that describe the same root-cause vulnerability, preserving every reporter's credit, every mailing-list thread reference, and every independent attack-vector description. Updates the kept issue's body in place, closes the duplicate with the duplicate label, and regenerates the CVE JSON attachment so both finders land in credits[].

Attempt to fix a security issue tracked in <tracker> by implementing the change in a public <upstream> PR. Runs security-issue-sync first to reconcile the issue's state, proposes an implementation plan, and on explicit user confirmation writes the change, opens a PR from the user's fork, and updates the <tracker> tracking issue. Public PR content is scrubbed so it does **not** reveal the CVE, the security nature of the change, or any link back to <tracker>.

Scan <security-list> for reports that have not yet been copied into <tracker> as tracking issues, present the proposed imports to the user, and — defaulting to *import unless the user rejects upfront* — create the tracking issues with the Needs triage project-board status and draft a receipt-of- confirmation reply to each reporter. This is the first step of the handling process: the entry point that converts an inbound email thread into a tracker the rest of the skills (security-issue-sync, security-issue-fix, generate-cve-json) operate on.

Open one or more <tracker> tracking issues from a markdown file containing a batch of security findings. Each finding becomes one tracker landing in the Needs triage board column. The file itself is the full report — there is no inbound reporter to reply to and no PR to inspect.

Open a tracking issue in <tracker> for a security-relevant fix that has already been opened (or merged) as a public PR in <upstream>, in the case where there is no inbound <security-list> report. The tracker lands in the Assessed board column with the scope label applied, pr created / pr merged reflecting the PR's state, and Remediation developer / PR with the fix body fields populated from the PR. Pairs with security-cve-allocate afterwards.

Triage a security scanner's multi-finding output (read via a pluggable scan-format adapter) and turn findings into security work only after a complete operator-reviewed triage. Reads the scan's finding index plus its per-finding evidence; buckets each finding by disposition; applies only the operator's confirmed per-entry decisions. Publishes the report as a gist and can open a report-back PR.

Optional sub-skill of security-issue-import, security-issue-invalidate, and security-issue-sync that handles the *relay/forwarder* case: a report that did not arrive directly from the reporter but was relayed onto <security-list> by an upstream broker (the ASF security team, a third-party disclosure platform, or an internal SOC). Runs after the parent skill's generic classification cascade, dispatches through adapters declared in forwarders.enabled per tools/forwarder-relay/README.md, applies the matched adapter's preamble-detect + credit-extract + reporter- addressing rules, and hands the routing decision back. Never mutates tracker state on its own.

Close an <tracker> tracking issue as invalid: apply the invalid label, remove the scope label, post a short closing comment, archive the item from the project board, and — for trackers imported from <security-list> — draft a polite-but-firm reply to the reporter on the original Gmail thread explaining the team's reasoning (extracted from the tracker's discussion). For trackers opened via security-issue-import-from-pr, the email-draft step is skipped per the *no outreach to the PR author* rule of that skill.

Synchronize a security issue in <tracker> with the state of its GitHub discussion, the <security-list> mailing thread, and any <upstream> PRs that fix it. The skill gathers all relevant signals and proposes label / milestone / assignee / field / draft-email updates — applying only what the user has explicitly confirmed. Suggests the next step in the handling process and prints the CVE allocation link when a CVE is needed.

For each open <tracker> issue carrying the needs triage label, read body + comments and classify the candidate disposition into one of six classes: VALID / DEFENSE-IN-DEPTH / INFO-ONLY / INVALID / PROBABLE-DUP / FIX-ALREADY-PUBLIC. On user confirmation, posts a triage-proposal comment that invites the security team to react. Read-only on tracker state — no label flips, closes, or CVE allocations. Supports --retriage for re-litigating passed-triage decisions when substantive new activity lands.

Generate a self-contained HTML dashboard of <tracker> repository statistics for security-team review.

Adopt and maintain the apache-magpie framework in a project repo via the snapshot-based installation mechanism. The only framework skill committed in an adopter's repo; every other skill is a symlink the install sub-action wires up. Sub-actions: /magpie-setup - first-time install (default; main-checkout only). adopt is an alias. /magpie-setup upgrade - refresh the gitignored snapshot per the committed lock (main-checkout only) /magpie-setup worktree-init - symlink a worktree's snapshot to the main's /magpie-setup verify - health check + drift detection /magpie-setup skill-sources - fetch/pin/symlink skills from trusted external sources listed in the adopter trust list (main-checkout only) /magpie-setup override <skill> - open or scaffold an agentic override in .apache-magpie-overrides/ /magpie-setup uninstall - reverse the install (snapshot, locks, symlinks, hook, doc sections); preserves .apache-magpie-overrides/ by default (main-checkout only). unadopt is an alias.

Probe the secure-agent setup for in-session functional restrictions that block legitimate workflows. Three live probes — SSH agent / Yubikey reachability, localhost port binding, docker / podman runtime socket — each pointing the user at the matching numbered troubleshooting entry and its settings.json remediation (see body). Read-only — never modifies settings.json, never invokes the sandbox bypass.

Guide an adopter through the first-time install of the framework's secure agent setup (bubblewrap + socat + claude-code, sandbox/permissions/clean-env layers). Walks every step interactively; never auto-runs sudo, shell-rc edits, or settings overwrites.

Surface drift between the user's installed secure agent setup and the framework's latest (framework checkout, pinned tools, user-scope script copies, denial commands, comdev MCP checkouts). Read-only — surfaces candidates and diffs, never auto-applies. The user decides what to update.

Walk the verification checklist for the framework's secure agent setup and report ✓ done / ✗ missing / ⚠ partial for each check, with concrete evidence (file paths, command output, version strings). Covers nine checks across settings wiring, installed tool versions, and sandbox configuration. Read-only — never modifies anything.

Walk an adopter through promoting a local .apache-magpie-overrides/<skill>.md file into a PR against apache/magpie. After the PR merges and the adopter runs /magpie-setup upgrade, the override file is no longer needed and the skill prompts for its removal.

Commit + push the user's shared Claude config to the ~/.claude-config private dotfile-style sync repo. Inspects for uncommitted local edits and unpushed commits, drafts a commit message, and after explicit approval commits and pushes. Runs git pull --rebase first if the local checkout is behind, so a push never overwrites concurrent work from another machine. Bootstraps the repo when it is missing: clones the default private remote if it exists, or creates a new private remote + scaffolds the minimal layout if it does not. Never force-pushes; never rewrites already-pushed history; never creates a public remote; never modifies files outside ~/.claude-config/ (except the confirmed fresh-host symlink wiring into ~/.claude/).

Show how the apache-magpie framework is adopted in the current repo, then adjust that setup in place. Renders a Markdown adoption dashboard: install method and pin, drift, the wired agent targets, the installed skill families, and symlink health. From the same view the user can add or drop agent targets and skill families; the actual change runs through the setup skill.

Turn a framework bug or quirk the agent hit while running a Magpie skill or tool into a fix PR against apache/magpie — one PR per issue. First confirms the problem is a framework defect (not a local misconfiguration or a stale snapshot), then searches apache/magpie for an existing issue or PR that already covers it, and only proposes a new fix PR when none exists — otherwise it points the user at the pending one.

Compare two near-duplicate skills — typically an ASF variant and a non-ASF or multi-project variant — and classify every difference as ALLOWED, DRIFT, or SAFETY-BASELINE. Produces a structured diff and a reconciliation proposal. Read-only: it never rewrites either skill; convergence is a separate confirmed authoring step. A safety-baseline divergence is always a must-fix, never silently merged into allowed-divergence noise.

Read-only GitHub Actions workflow security audit for one repository, an explicit repository set, or a whole GitHub org. Runs zizmor to surface injection vulnerabilities, excessive permissions, unpinned external actions, and self-hosted-runner fork-secret leaks. Produces a grouped, prioritised finding report; never edits workflow files, opens PRs, or posts comments.

Author a new skill for the Apache Magpie framework, or update an existing one. Walks the user through the framework's skill shape (frontmatter, resources, placeholder convention, prompt-injection defences, Privacy-LLM gate-check) and validates via the framework's existing [tools/skill-and-tool-validator](../../tools/skill-and-tool-validator/). Scaffolds new skills via init_skill.py.