Menu
Help a CompeteHub member regain enough context to continue assigned work.
Installer avec Codex ou Claude Copiez ce prompt, collez-le dans Codex, Claude ou un autre assistant, puis laissez-le vérifier la page du skill et l'installer pour vous.
Help the human owner run CompeteHub delivery planning, issue slicing, PR readiness review, merge-readiness analysis, and standup sync.
Route non-trivial CompeteHub work through the project delivery workflow. Use when a task touches product scope, architecture, implementation, testing, documentation, GitHub issues or PRs, course reports, or agent-assisted delivery decisions.
Use CompeteHub's TDD workflow for testable behavior changes, bug regressions, backend rules, API validation, auth/permission checks, state transitions, reminders, subscriptions, recommendations, and review/publication workflows.
Deliver one focused CompeteHub implementation, bug fix, test, documentation, or process slice from issue or objective through impact analysis, TDD decision, code/docs changes, validation, and PR-ready summary.
Toolkit for interacting with and testing local web applications using Playwright. Supports verifying frontend functionality, debugging UI behavior, capturing browser screenshots, and viewing browser logs.
Basé sur la classification professionnelle SOC
| name | member-help |
| description | Help a CompeteHub member regain enough context to continue assigned work. |
| disable-model-invocation | true |
Member Help is a user-invoked, read-only help-forward skill. Use it only when a
collaborator explicitly asks for /member-help.
The goal is not to diagnose once and stop. The goal is to help the agent and member share enough context to continue assigned project work.
Confirm the member, assigned work, current intent, and blocker.
Prefer member-provided assignment context:
If the member only provides identity and no assigned work, ask for the standing meeting issue or owner list. If that is unavailable, proceed with assignment-recovery and state that the result is only a candidate, not a new assignment.
Completion criterion: member identity and assignment source are known, or the missing assignment is explicitly marked as recovery.
Use the current conversation context when it is sufficient. Do not re-index just because this step exists.
If context is insufficient, run a lightweight index pass. Read metadata first, not full issue bodies, comments, PR diffs, or CI logs.
Useful index fields:
Completion criterion: the likely issue/PR/docs candidates are narrowed to the smallest useful set.
Deep-read only what is needed for the member's next step.
Default hydrate set:
Read comments, diffs, review threads, or CI logs only for review/debug/blocker work.
Completion criterion: the agent can explain the assigned work, relevant constraints, and next action without guessing.
Continue the member toward work. Do not produce a long diagnostic report by default.
Common help-forward outcomes:
delivery-slice, tdd, review, or another workflowIf the member explicitly asks the agent to implement, edit, commit, create an
issue, open a PR, or perform external writes, stop Member Help and switch to the
appropriate project workflow. Follow AGENTS.md and preview required writes.
Completion criterion: the member and agent have enough shared context to keep working in the current conversation or enter a concrete workflow.
Read references/patterns.md only when a concrete help pattern or reusable
member prompt would improve the response.