Ensures sprint tickets meet strict acceptance criteria before moving them to the Done state.
Install
mkdir -p .claude/skills/metravel-sprint-reviewer && curl -L -o skill.zip "https://agentskills.codes/api/skills/download/13365" && unzip -o skill.zip -d .claude/skills/metravel-sprint-reviewer && rm skill.zipInstalls to .claude/skills/metravel-sprint-reviewer
Activation
This is the description your AI agent reads to decide when to run this skill — the better it matches your request, the more reliably it fires.
Accept or reject metravel task-board tickets in an active sprint using real Done-gate evidence. Use when Codex is asked to review/close sprint tickets, move tasks from review/testing to done, verify Task Contracts, run acceptance checks, or decide what can safely ship from the MCP task board.Key capabilities
- →Load the active sprint or a requested task
- →Read task descriptions, dependencies, blockers, and Task Contracts
- →Reject tasks with refinement gaps before runtime work
- →Run or inspect validation required by the Done gate
- →Append evidence to the task description
- →Move passing tasks to 'done' or return tasks to 'in_progress' with defects
How it works
The skill loads tasks, reviews their details and contracts, performs validation based on Done gate requirements, and updates task statuses with evidence.
Inputs & outputs
When to use metravel-sprint-reviewer
- →Accepting sprint tickets
- →Verifying task completion
- →Reviewing pull request evidence
About this skill
Metravel Sprint Reviewer
Use this skill for board acceptance, not implementation. Code fixes go back to the owning implementation skill.
AGENTS.md is inherited. Read the ticket's Task Contract, the matching feature
contract, and only the acceptance/status/evidence headings from
docs/TASK_BOARD_MCP.md required by that gate.
Board Rules
- Use the shared MCP task board tools only.
- If board endpoints return
HTTP 401, refresh the staff token through.env.e2eusingdocs/TASK_BOARD_MCP.md; never print the token. - Review only the requested ticket or active sprint scope.
- Review
area=backtickets only when the user explicitly requests backend acceptance/status work. Otherwise skip them and report the skipped count. - Do not create new feature code while acting as reviewer.
- Do not move a task to
donewithout real evidence for its Done gate. A live shared gate records onlyvalidation delegated: active gate pid/name; it is notpassed. Request its result and resume acceptance instead of closing or parking the task.
Acceptance Loop
- Load the active sprint or requested task.
- For each candidate in
testing, read the full description, dependencies, blockers, and Task Contract. Any ticket inreview,todo,in_progress, orbacklogis not an acceptance candidate: route it to its owning stage and do not run browser/API/device QA. - Reject refinement gaps before runtime work:
- missing
Scope,User-visible result,Data/API contract,Platform impact,Localization impact,Dependencies,Fallback/mock policy,Validation,Regression control, orDone gate Regression controlempty ornoneon akind=bugor consolidation task- unresolved blocking dependency
needs_human=truewithout the human step completed
- missing
- Run or inspect the exact validation required by the Done gate. Reject evidence that
violates
docs/TASK_BOARD_MCP.md→#### Качество evidence:- a
200/no-crash result offered for a task about size, count, duration or order, with no before/after number - no negative probe where an unsupported input exists: it must be observably distinguishable from a valid one, never a silent heavy/generic fallback
- a build-time or deploy-time check offered for a user/crawler-visible production surface instead of a recurring production probe
production-backed preview, local production build, production API behind a local bundle, or deploy logs offered as evidence that the live production URL is fixed- a production optimization without the same before/after URL, viewport/browser/DPR, auth/cache state, request/API count, bytes, response codes, and relevant scroll/lazy-loading probe
- a page-level performance task whose evidence proves one component or URL but does not remeasure the whole page budget and adjacent consumers touched by shared media/source/pagination code
- a test that mocks the very primitive under investigation offered as contract evidence
- a consolidation closed without naming the CI guard that fails when the new single contract is bypassed
- when the required automated gate is already owned by another live session, apply the coordination contract from
AGENTS.md: do not duplicate it; request the result/unblock action and resume this acceptance pass Then collect every applicable positive check required by the contract: targeted tests or governance checks, browser/API probes against the target environment,npm run test:i18nplus affected locale evidence, and device/mobile evidence only when a frontend ticket owns Android- or iOS-specific observable behavior. Client/device evidence belongs to a linked frontend task, not to a backend Done gate.
- a
- Append evidence to the task description without erasing prior history.
- Finish each started acceptance with a decision. Passing task →
done. Unfinished work promised by the current ticket →todo, orin_progressonly when that work actually starts. A separate confirmed defect → run$metravel-problem-memory, create/reuse a linked task, and do not park the accepted ticket. Keeptestingonly for an exact repeated measurement or temporal gate with parameter, threshold, current value, trigger/earliest recheck, and command/scenario. Missing access/device/gate result stops the turn for an unblock request; it is not a board verdict. Useblocked_byonly when a hard dependency prevents implementation from starting or continuing.
Backend Acceptance
For an explicitly requested area=back review, accept only the evidence that is
both relevant to the backend task and available to this workspace: read-only
origin/master source inspection, exact API/HTTP probes, production/runtime
status, database/log/queue observations when access exists, and negative or
authorization probes where the contract requires them.
- Do not require Android or iPhone evidence for a backend ticket. A client flow
that consumes the API is verified in its linked
area=fronttask. todomeans backend implementation, refinement, deploy, configuration, data, or other owner work still remains. Return a failed backend ticket totodowith the exact observed defect and required owner action.- Keep a completed backend ticket in
testingwhen a concrete, executable in-scope time window or observation period has not elapsed yet. Record the parameter, threshold, current measured value, earliest recheck/trigger, and exact probe; do not return it totodowhen there is nothing to implement or operate. - Move the ticket to
donewhen backend work is complete and all available, relevant mandatory source/API/production probes are green. Irrelevant, unavailable, or client/device evidence outside backend ownership does not block acceptance and must not create a synthetictodoitem.
For a production incident, no deploy authorization is not a reason to lower the
Done gate. If the live probe is mandatory, request exact deploy authorization
and stop the acceptance decision until it is granted; resume with the real
remeasurement. Do not describe production as fixed, close, or park the task from
local fix ready alone.
Output
Return a compact Sprint Review:
- tasks accepted with evidence
- tasks rejected with blockers
- checks/probes run
- tasks that need human or backend follow-up
- any board/token/access blocker
When not to use it
- →The task involves implementation or code fixes
- →The task is not related to board acceptance
- →The task requires creating new feature code
Limitations
- →This skill is for board acceptance, not implementation
- →It does not create new feature code
- →It only reviews the requested ticket or active sprint scope
How it compares
This workflow systematically validates task completion against defined criteria and evidence, unlike a manual review that might overlook specific Done-gate requirements.
Compared to similar skills
metravel-sprint-reviewer side by side with the closest alternatives in the catalog.
| Skill | Installs | Updated | Safety | Difficulty |
|---|---|---|---|---|
| metravel-sprint-reviewer (this skill) | 0 | 3mo | Review | Intermediate |
| antigravity-workflows | 0 | 5mo | No flags | Intermediate |
| gsd | 0 | 5mo | Review | Advanced |
| agent-project-board-sync | 0 | 7mo | Review | Intermediate |
Try saying
Example prompts that trigger this skill in your AI assistant.
More by kelios
View all by kelios →You might also like
antigravity-workflows
gabryssad9-del
Orchestrate multiple Antigravity skills through guided workflows for SaaS MVP delivery, security audits, AI agent builds, and browser QA.
gsd
dturkuler
Antigravity GSD (Get Stuff Done) - A spec-driven hierarchical planning and execution system. Triggers on project planning, phase management, and GSD slash commands.
agent-project-board-sync
ruvnet
Agent skill for project-board-sync - invoke with $agent-project-board-sync
linear-ci-integration
jeremylongshore
Configure Linear CI/CD integration with GitHub Actions and testing. Use when setting up automated testing, configuring CI pipelines, or integrating Linear sync into your build process. Trigger with phrases like "linear CI", "linear GitHub Actions", "linear automated tests", "CI linear pipeline", "linear CI/CD".
closeout
LaPaGaYo
|
regression
ThePalaceProject
Run a full release regression test — sets up workspace, runs automated tools, guides manual testing, generates report and Jira tickets. Use for release gates and QA cycles.