Help us improve
Share bugs, ideas, or general feedback.
npx claudepluginhub gsaranti/pantheon --plugin metisHow this skill is triggered — by the user, by Claude, or both
Slash command
/metis:metis-implement-taskThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Implement the task described by the plan in context, and close with a scope report.
Guides technical evaluation of code review feedback: read fully, restate for understanding, verify against codebase, respond with reasoning or pushback before implementing.
Share bugs, ideas, or general feedback.
Implement the task described by the plan in context, and close with a scope report.
A plan or description present in conversation context. It can come from any of:
/metis-plan-task invocation in this session — structured plan in chat history..metis/CURRENT.md loaded by /metis-session-start — a prior session captured the plan there.If none of these are present, ask the user before proceeding. Implementing against an unstated brief is the silent-drift failure mode this skill exists to avoid.
.metis/CONTRADICTIONS.md, .metis/QUESTIONS.md, .metis/RESOLVED.md.references/honest-scope-reporting.md — read before composing the closing scope report.
.metis/. The plan, the build spec, session state — none of these are implementation-time concerns. /metis-session-end is what captures implementation state in .metis/CURRENT.md if the session ends before review..claude/.Trailing prompt: see ${CLAUDE_PLUGIN_ROOT}/references/command-prompts.md.
If the verification command fails repeatedly, prefer surfacing the problem to grinding on it. Three signals that the right move is to stop and report rather than keep fixing:
When any of these hit, stop. Do not commit further fixes. Close with the verification output as-is, a one-line description of what was tried, and a Finding flagging the likely failure mode (description-vs-implementation gap, stale test fixture, dependency the plan didn't anticipate, two unrelated outcomes bundled into one task, etc.). A verification failure that takes the implementation outside its planned shape is a signal that the plan or description was wrong — not a problem the implementer should solve by widening on the fly.
Before returning:
references/honest-scope-reporting.md — four categories (Skipped / Deferred / Stubbed / Handled differently), no defense. An empty report is fine when the work met its criteria cleanly; say so in one line.The scope report and verification output live in the return message in chat. /metis-session-end captures them in .metis/CURRENT.md if the session ends before review.
/metis-review-task to dispatch the reviewer subagent against the diff + scope report.