CtrlK
BlogDocsLog inGet started
Tessl Logo

subagent-execution-governance

把子代理派发从"监督型治理"升级为"契约型治理":探查/实现/验证角色分离、契约快照(含负空间)、三级有界读取、单写者文件所有权、行动门槛替代轮次门槛、结构化 BLOCKED、token/工具/轮次预算、结构化结果回收与 repair-contract 恢复。用于派发实现型子代理(写代码/写文件类任务)、设计"谁读什么/谁写什么/何时算阻塞"的执行协议、审计子代理空转(反复探查不产出、并行写冲突、上下文膨胀)、或复盘"子代理烧了大量 token 却没产出"的会话。不用于纯问答、单轮小修改、或已有成熟任务流程的普通委派。

71

Quality

87%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

SecuritybySnyk

Passed

No findings from the security scan

SKILL.md
Quality
Evals
Security

Quality

Content

75%Weight 40%Scale 1-5

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

A dense, highly actionable governance skill: concrete templates, named budgets, state machines, and real tooling with invocation syntax, plus explicit validation gates and a repair loop. The main weaknesses are some recurrence of the same audit figures across sections and inlined plugin/template detail that would scan better as one-level-deep reference files.

Suggestions

Consolidate the recurring b21b9211 / 309万-token / 169-injection figures into a single 反面案例 section and reference it, instead of restating them in the preamble, section 九, and P2 notes — this trims the conciseness gap toward 5.

Move the P1/P2 plugin behavioral specs and the contract-snapshot / BLOCKED field templates into references/ files (e.g. contract-template.md, blocked-schema.md) and link them one level deep, improving progressive_disclosure and navigation.

Add one explicit end-to-end orchestration procedure (ordered steps with checkpoints) that stitches the twelve sections into a single dispatch→contract→implement→validate→recover flow, pushing workflow_clarity from 4 to 5.

DimensionReasoningScore

Conciseness

Largely lean and assumes Claude's competence — the role/read/write table, contract template, and state machine are token-dense but information-dense, not explanatory padding. Held below 5 by occasional repetition of the same anti-pattern (the b21b9211 / 309万token figures recur across the preamble, 反面案例, and P2 sections) that could be consolidated.

4 / 5

Actionability

Concrete, copy-paste-ready artifacts throughout: the contract-snapshot field block, the BLOCKED eight-field schema, the three-level read table, the seven-item dispatch checklist, and named scripts with invocation syntax (session-discipline-audit.js <session id>). Falls short of 5 because the P1/P2 plugin specs describe behavior rather than giving runnable install/registration commands.

4 / 5

Workflow Clarity

Clear multi-step sequence (Discovery→Contract→Implementation→Validation) with explicit validation gates (BLOCKED_INVALID gate, file-conflict intersection check, action-threshold mutation/BLOCKED gate) and a recovery loop (FAIL→repair contract→re-dispatch). Not 5 because the end-to-end orchestration sequence across the twelve sections is implied rather than given as one explicit ordered procedure with checkpoints.

4 / 5

Progressive Disclosure

Well-structured overview (12 numbered sections + checklists + reverse-case table) with one-level-deep, real bundle references (scripts/session-discipline-audit.js, scripts/splice-ab-experiment.js both present and invoked with syntax). Held below 5 because the P1/P2 plugin internals and the contract/BLOCKED templates are inlined at length where a references/ file split would aid navigation.

4 / 5

Total

16

/

20

Passed

Description

100%Weight 40%Scale 1-5

Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.

An exceptionally strong description: specific concrete mechanisms, natural user-voiced trigger phrases, explicit what/when/when-not, and a clearly bounded niche with anti-trigger guidance. It is long, but every clause earns its place and the verbosity is justified by the precision rather than padded.

DimensionReasoningScore

Specificity

Lists multiple concrete actions with comprehensive coverage: 探查/实现/验证角色分离, 契约快照(含负空间), 三级有界读取, 单写者文件所有权, 行动门槛替代轮次门槛, 结构化 BLOCKED, token/工具/轮次预算, 结构化结果回收, repair-contract 恢复 — well beyond the 4-anchor bar of 'several specific actions; minor gaps'.

5 / 5

Completeness

Explicitly answers both 'what' (the upgrade from 监督型 to 契约型 governance with the enumerated mechanisms) and 'when' (用于...派发实现型子代理/设计执行协议/审计空转/复盘), and adds explicit anti-triggers (不用于纯问答、单轮小修改、已有成熟流程的普通委派) — the strongest anchor with concrete trigger phrases.

5 / 5

Trigger Term Quality

Comprehensive natural phrasing a user would actually say: '派发实现型子代理(写代码/写文件类任务)', '审计子代理空转(反复探查不产出、并行写冲突、上下文膨胀)', '复盘 子代理烧了大量 token 却没产出 的会话' — covering synonyms and the literal failure modes that trigger this skill.

5 / 5

Distinctiveness Conflict Risk

Clear niche (契约型 governance for implementation-type subagent dispatch) with distinct triggers and explicit not_for boundaries that separate it from adjacent execution-discipline/decision-gates skills; minimal conflict risk.

5 / 5

Total

20

/

20

Passed

Validation

93%

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation15 / 16 Passed

Validation for skill structure

CriteriaDescriptionResult

frontmatter_unknown_keys

Unknown frontmatter key(s) found; consider removing or moving to metadata

Warning

Total

15

/

16

Passed

Repository
ooooooooooooooooooop/agent-tools
Reviewed

Table of Contents

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.