Skill Claude CodeCodex
A release workflow for repositories that contain multiple plugins, known as a monorepo. It audits the plugins, runs tests, fixes discovered problems, updates plugin versions, and creates a repository tag.
Skill Claude CodeCodex
Set up p-chat in the current repo — walk the owner through creating a Telegram bot and a token file, then run pchat init to verify the token, discover the chat id, and write .pchat.json. Use when the user says "init p-chat", "set up telegram chat", or "connect telegram".
Skill Claude CodeCodex
Answer pending free-text Telegram questions via the pchat CLI — read pending questions, answer each grounded in the repo, send, ack, and append to the session transcript. Use as the prompt of a p-shed chat-responder job, or when the user says "answer the chat" / "check telegram questions".
Skill Claude CodeCodex
Use after the plan is approved and you are about to implement it. Drives the steps in order — one at a time — invoking test-driven-development for code steps and verification-before-completion after each, marking a step done only when its acceptance criterion is met. Steps live in plan.md ## Steps (legacy) or, when…
Skill Claude CodeCodex
Use when processing review feedback — a review follow-up (a p-tasks sub-task in canonical mode, or a ## Review follow-ups item in plan.md in legacy mode), a PR comment, an inline reviewer message. Enforces verify-the-finding-first discipline before implementing or rejecting. Counterpart to requesting-code-review.
Skill Claude CodeCodex
Use after verification-before-completion passes and there is a diff worth reviewing. Dispatches a code-review subagent (via Task tool with general-purpose + inline template) on the branch diff, then leads the user through severity-aware triage and records accepted findings as follow-ups — p-tasks sub-tasks in…
Skill Claude CodeCodex
Use after verification-before-completion passes and the implementation should be checked against the spec. Dispatches a task-review subagent (via Task tool with general-purpose + inline template) with paths to spec, feature.feature, adr.md, the branch diff (plus plan.md in legacy mode). Records accepted findings as…
Skill Claude CodeCodex
Use to execute an approved plan in the current session by dispatching a fresh implementer subagent per step, a per-step review (spec compliance + code quality) after each, and a broad whole-branch review at the end. The isolated alternative to executing-plan — fresh context per step, artifacts handed over as files so…
Skill Claude CodeCodex
Use when verification fails, a test goes red, or behaviour is unexpected — before proposing any fix. Enforces reproduce → hypothesise → test the hypothesis → narrow → fix at root cause → re-verify, one hypothesis at a time. Stops guess-and-check debugging.
Skill Claude CodeCodex
Use when starting a new non-trivial task to elicit requirements through dialog and produce specs/ /specification.md (always) plus optional feature.feature and adr.md. Invoked by /p-flow:task-start. Hard gate — does not invoke writing-plan or any implementation skill until the user approves the written spec.
Skill Claude CodeCodex
Exit point of the p-flow task development flow. Runs pre-checks (no main branch, no uncommitted changes, plan steps checked, verification marker fresh), pushes the branch, and prints ready-to-copy MR creation commands for both GitHub (gh) and GitLab (glab). Never runs gh or glab itself. Usage /p-flow:task-end.
Skill Claude CodeCodex
Entry point to the p-flow task development flow. Resolves slug + branch type + optional worktree, runs all conflict checks BEFORE any side effects, then atomically creates the branch (and worktree), opens specs/ /, and invokes task-brainstorming. Usage /p-flow:task-start [--worktree].
Skill Claude CodeCodex
Use when about to write code (production code path, not docs/scripts/research) — enforces RED-GREEN-REFACTOR — write a failing test first, watch it fail, write minimal code to pass, verify pass, refactor. Pairs with verification-before-completion (this skill is the "before code" gate; verification is the "before…
Skill Claude CodeCodex
Use when starting work that benefits from isolation from the current checkout (parallel feature branches, long-running experiments, freeing the main checkout for unrelated work). Reference documentation for safe worktree creation, common pitfalls, and cleanup.
Skill Claude CodeCodex
Use when starting any conversation in a repo with p-flow enabled — establishes the p-flow task development flow surface (commands, skills, when to invoke each) so the model can pick the right tool without keyword guessing.
Skill Claude CodeCodex
Use before claiming any task is done, fixed, ready, implemented, or before any commit. Runs detected tests/lints, quotes concrete output, and writes a state marker at .claude/.p-flow-state/ /last-verification so /p-flow:task-end can detect that verification ran. Evidence before assertions, always.
Skill Claude CodeCodex
Use after a spec exists at specs/ /specification.md to produce a step-by-step implementation plan — as p-tasks sub-tasks when p-tasks is present (no plan.md), or specs/ /plan.md when it is absent. Refuses to write a step without an acceptance criterion. Decomposes work into 5–15 steps; flags larger work for sub-task…
Skill Claude CodeCodex
Use when creating a new p-flow skill or substantially editing an existing one — establishes the conventions (frontmatter shape, section order, dispatch patterns, template placement, test coverage) so the plugin stays internally consistent.
Skill Claude CodeCodex
Show the pgraph command cheat-sheet and when to use each. Use when the user says "p-graph help" or asks what pgraph can do.
Skill Claude CodeCodex
Answer a structural question about the codebase from the code graph, with concrete file:line citations. Use for "who calls X", "what breaks if I change Y", "how does X reach Y", "where is X defined", "explain symbol X", "ask the graph", "query the code graph".
Skill Claude CodeCodex
Explicitly rebuild the p-graph code graph — full or incremental. Day-to-day freshness is automatic (queries auto-refresh); use this for a full rebuild after a big refactor, or to warm the graph after a pull. Use when the user says "sync p-graph", "reindex", "rebuild the code graph", or "full reindex".
Skill Claude CodeCodex
Add, modify, or delete a scheduled job in .pshed/jobs.yml (cron schedule + folder + prompt). Use when the user says "add a job", "schedule a run", "change the schedule", "disable a job", or "delete a job".
Skill Claude CodeCodex
Un-stick a job that p-shed stopped scheduling — clears the process-level circuit breaker and the task-level self-pause marker (an operator pause is lifted with pshed resume --id instead). Use when the user says "reset the breaker", "un-stick a job", "my job stopped running", "resume a paused job", or after a job…
Skill Claude CodeCodex
Install the OS scheduler entry that runs pshed tick every minute in this folder (Windows schtasks, Linux/macOS crontab). Use when the user says "start p-shed", "enable the scheduler", or "begin running jobs".