跳到正文
cccskills
免费在 GitHub 公开

sparc-review

Audit an existing SPARC run without changing it when the user asks for gate readiness, trace integrity, evidence coverage, or an independent delivery review.

中文简介正在准备中。当前显示原文说明。

查看安装方法

包含的文件(2)

  • SKILL.md2.9 KB
  • agents/openai.yaml396 B

SKILL.md(原文)

安装前可以查看交给 AI 代理的指令内容。

SPARC review

Treat the review as read-only. Use only sparc_run_get, sparc_phase_get, sparc_gate_validate, and sparc_trace_get from the sparc MCP server.

MCP prerequisite

Before the first tool call, verify that the host exposes an MCP server named sparc with the documented eight sparc_* tools. The standalone npm skill installer copies skills only and intentionally does not edit host MCP configuration. If the server is missing, stop and ask the user either to install the SPARC plugin or explicitly register this exact pinned stdio launcher:

npx --yes @ruvnet/sparc@1.0.0 mcp stdio

Do not substitute a floating package tag, silently edit .mcp.json, .openai.mcp.json, or user-level host configuration, or claim the skill is operational until the server is visible. ChatGPT requires the separately deployed authenticated HTTPS transport described by SPARC; it cannot use this local stdio command.

Review method

  1. Call sparc_run_get and record its summary revision, digest, current phase, counts, and receipt tail. It is a summary, not the full run.
  2. For each entered phase, follow every sparc_phase_get pagination.nextCursor until absent. Follow every sparc_trace_get page the same way. Distinguish an artifact's presence from its gate passing.
  3. Validate the current gate. Check phase order and trace integrity; require each usable pass or exception to report verified Ed25519 attestation. Confirm Refinement and Completion cite the exact immutable { evidenceId, version, digest } returned for the same requirement and registered acceptance test.
  4. Classify findings as blocker, risk, or observation. Cite the affected phase, artifact or evidence identifier, and the unmet gate rule.
  5. End with a clear pass or fail decision and the smallest remediation that would satisfy each blocker.

Treat cursors as opaque snapshot tokens. If a cursor becomes stale or a reread shows a changed revision or digest, discard every collected page and restart; never combine pages from different snapshots. Read the summary again before reporting.

An attestation proves that a public key configured through SPARC_EVIDENCE_VERIFIER_KEYS verified the exact evidence subject, including the run's immutable genesisDigest and requirementsDigest; confirm those stored bindings match a fresh run summary. It does not prove that the underlying result is true or that signer custody is independent. Report absent verifier provenance or shared writer and signer control as a risk even when the cryptographic check passes.

Do not submit artifacts, record evidence, advance phases, run commands, modify files, request verifier private keys, or turn an exception into a silent pass.

评价

还没有评价。欢迎分享你的使用感受。

评论

无需登录即可发表。欢迎分享使用感受、提问,或说明在哪些环境下无法运行。

还没有评论。来写第一条吧。

提交后会立即公开。请勿填写邮箱、电话等个人信息。URL 最多 2 个。

0 / 1000 字

此表单受 reCAPTCHA 保护,适用 Google 的隐私权政策和服务条款。

同一仓库的技能

sparc

免费

Run an engineering goal through the gated SPARC phases when the user wants a new implementation, integration, redesign, or other substantial delivery workflow.

中文简介正在准备中。当前显示原文说明。

ruvnet/sparc4782026年10月17日 更新

Safely resume an interrupted or handed-off SPARC run from its verified persisted state without duplicating mutations or skipping unfinished gates.

中文简介正在准备中。当前显示原文说明。

ruvnet/sparc4782026年10月17日 更新

查看 ruvnet 的全部技能

报告此技能的问题