ask-session-review
Session Review: Reflects on skill usage and gaps and files GitHub issues from bugs, findings, or follow-ups. Use after completing work. Common triggers: retrospective, retro, reflect on session, how did i use skills, file an issue, create issue, github issue, file issue, gh issue create.
- License 许可
- MIT
- 最近更新
- 2026-11-09
- Stars
- ★ 1
- Forks
- 0
01 前言 Front Matter
- name
- ask-session-review
- description
- Session Review: Reflects on skill usage and gaps and files GitHub issues from bugs, findings, or follow-ups. Use after completing work. Common triggers: retrospective, retro, reflect on session, how did i use skills, file an issue, create issue, github issue, file issue, gh issue create.
- disable-model-invocation
- true
02 资源文件 1
- 脚本check-existing-issue.sh451 B
03 SKILL.md 正文
ASK Session Review
Two modes: session self-review (primary) and general issue filing (fallback).
Session Review Mode
Use at session close to evaluate whether the skill ecosystem served the work well.
This mode reviews only ASK skills supplied by the active Agent Skills Kit distribution. A candidate is ASK-owned when it was loaded from its canonical ASK source: the skills/*/SKILL.md tree in a source checkout or the host's installed/shared ASK skill root. Do not infer ASK ownership from its mention in a consuming repository's instructions.
- Identify ASK skills loaded during the session from
matchedSkillsin session state. - For each ASK skill, check: did it trigger at the right time? Was its guidance complete? Was anything missing?
- Do not treat a project-specific skill named in a consuming repository's instructions as an ASK skill. Its availability and guidance belong to that project's source repository; do not create an ASK issue or set
shouldCaptureImprovementfor it. - If an ASK gap, missing trigger, or improvement opportunity emerges → create issue in
MarkBovee/agent-skills-kit. - Issue title starts with
skill:and names the ASK skill plus the gap (e.g.skill: session-review lacks triggers for self-review). - Body includes: observed pattern, why current behavior falls short, proposed change.
- Before filing, check for equivalent open issues with
gh issue list --searchagainstMarkBovee/agent-skills-kit. - If an ASK improvement is confirmed, flag
shouldCaptureImprovement: truein session state.
For a project-specific skill gap, report the owning project and repository in the review output. File an issue there only when the user explicitly requests general issue filing mode.
When code changes exist, report stale or mismatched review evidence explicitly, including expected edit generation or diff identity when available. Do not leave an unresolved review obligation looking complete.
When session state is unavailable
Reconstruct relevant skill use from the transcript: explicit skill calls, loaded SKILL.md files, router matches, and the work actually performed. Mark this reconstruction as best-effort. If state cannot be written, record the improvement in the review output or issue instead of claiming the session flag was updated.
Record release-gate cost
When the session ran a release gate, record what the validation/audit actually cost versus what was planned (see intake's release-gate cost decision). Note the outcome: a delta re-check that a standard tier closed in seconds, or an over-scoped deep run that wasted minutes and was aborted. List the actual cost per gate (reviewer, first audit, delta audit: wall time and tokens when known) next to the planned Gate cost: line, and name any gate that found a defect the previous gate missed. Feed that back as skill: guidance so the next release starts at the right tier instead of repeating the expensive default.
Evidence-aware issue communication
When writing an issue, follow the shared evidence-aware policy:
- Establish current state from the complete available context before describing a gap.
- Distinguish observed evidence, conclusions, and information still needed.
- Keep the workflow phase explicit when a report has moved into implementation, release, or verification.
- Ask for new evidence only when a changed version, configuration, or environment makes a fresh capture necessary; explain what it will verify.
- Prefer the smallest concrete next step over a broad diagnostic checklist.
Issue Filing Mode
For explicit issue creation requests (bug report, review finding, follow-up):
- Infer repo from current git remote or explicit
owner/repo. - Check
ghauth before usinggh issue create. - Check for duplicates with
gh issue list --searchor the bundled helper .github/skills/ask-session-review/check-existing-issue.sh. - Build compact body: problem, impact, repro/evidence, expected outcome.
- Create issue directly — no draft approval loop.
- Return URL and note assumptions.
Use with
code-reviewwhen review findings should become follow-up issueswrite-skillwhen improvement should become a tracked changeverificationwhen session is closing and reusable learning should not get lost
Avoid
- Filing issues for one-off typos or isolated repo bugs
- Guessing a repo outside the current checkout
- Duplicate issues when a clear match exists
- Vague meta-issues without observed pattern or proposed change
04 同仓库更多 Skills
ask-write-skill
Write Skill: Creates and revises skills and guardrails from reusable workflow gaps. Use when making a skill or when agent behavior exposes a routing gap.
ask-verification
Verification: Proves work is fixed or ready and cleans up changed repos. Use before claiming something works or handing off.
ask-text-writing
Text Writing: Writes emails, posts, articles, bios, READMEs, and docs that sound human and avoid AI writing patterns. Use when drafting or rewriting prose.
ask-summary
Summary: Explains what changed and what is proven, walks through code, or re-pitches unclear output. Use when work is done or the user asks what changed.
ask-session-review
Session Review: Reflects on skill usage and gaps and files GitHub issues from bugs, findings, or follow-ups. Use after completing work.
ask-research
Research: Does bounded fact-finding across the repository, docs, code, or external sources. Use when a question needs evidence before an answer or decision.