[cao-evolution:compiler-security] Compiler security scan: one control.md shellcheck fix; all other findings upstream-owned (gh-aw, container images)

クローズ 初心者向け
  • 何か問題がありますか？
#17,475 コメント 0 件 リアクション 0 件 担当者 0 名 GitHub で見る

メンテナーはふだん 1 日以内に返信

まだ誰も着手していません。

評価

難易度
1/5
見積もり時間
1〜3時間
初心者へのやさしさ
62/100
issue の種類
バグ
明瞭さ
明確に書かれている
活発さ
活発
技術スタック
github-actions, javascript, shell
領域
devops, security

調査の方向性

.github/workflows/shared/control.mdを読み、3行の出力によってshellcheck SC2129が発生するcao_target_read_scopeステップを見つけてください。issueの説明に従って、1つのリダイレクトの下に各行をまとめ、その後workflowをコンパイルし、対象に割り当てられた指摘が解消されたことを確認してください。issueでは、生成されたロックファイルの差分を確認し、マージ前に完全な検証とセキュリティスキャンを再実行することが求められています。

索引モデルが issue の本文から書いたものです。

説明

cao-evolution cao-evolution:compiler-security

githubnext/gh-aw-cao compiled cleanly for structure (70/70 workflow sources → 70/70 lock files) but the full gh-aw validation/security suite exited non-zero (exit code 1). Of the ~50,000 lines of tool output, every supported finding traces to the upstream github/gh-aw compiler (generated boilerplate shared by all 70 workflows, including unresolved daily_ai_credits_*/safe-outputs-app-token template properties and compiler-selected container images/CLI installs) or to third-party base images (alpine:latest, Ubuntu/bind9/Python packages in gh-aw-firewall images, golang.org/x/net/Go stdlib CVEs, npm tar/http-cache-semantics advisories) and their licenses. Exactly one target-owned actionable finding exists: a trivial shellcheck style note (SC2129, not a security issue) in the repeated cao_target_read_scope step defined in .github/workflows/shared/control.md, which all 70 workflows import. Grant license-policy evidence is available (.grant.yaml present) and flags upstream image package licenses only — no target license-policy gap. Highest severity among supported tool output is Critical (Go stdlib/golang.org/x/net CVEs in the gh-aw-firewall/agent image), but it is entirely upstream-owned. Recommended next action: apply the one-line shellcheck style cleanup in control.md; no other target remediation is warranted from this run.

Action: Assign this issue to Copilot using Agent prompt below; review its pull request and merge only after rerunning the full compiler and security scan confirms that the target-owned actionable findings are resolved.

Failure details
  • Target repository: githubnext/gh-aw-cao
  • Compiler exit code: 1
  • Workflow sources checked: 70
  • Generated lock files checked: 70
  • Result: findings
  • Grant license-policy evidence: available
  • Target-owned actionable findings: 1
  • Upstream-owned context findings: many (deduplicated below)
Findings table
Tool Workflow/Image Owning component/repository Ownership Severity Finding Target remediation
shellcheck (actionlint) shared/control.md → all 70 .lock.yml githubnext/gh-aw-cao (control.md) target-owned actionable style (SC2129) cao_target_read_scope step uses 3 sequential >> "$GITHUB_OUTPUT" redirects instead of a single grouped block Combine the three echo ... >> "$GITHUB_OUTPUT" lines into { ... } >> "$GITHUB_OUTPUT" in control.md
actionlint (expression) all 70 .lock.yml github/gh-aw compiler upstream-owned context error needs.activation.outputs.daily_ai_credits_total/threshold/guardrail_status/exceeded/estimated referenced in env blocks but never declared in the activation job's outputs: map (compiler template/output-wiring bug) None — report to github/gh-aw
actionlint (expression) self-care-dashboard-performance.lock.yml, self-care-pages-health.lock.yml github/gh-aw compiler upstream-owned context error steps.safe-outputs-app-token.outputs.token referenced but that step id is absent from the generated job None — report to github/gh-aw
zizmor all 70 .lock.yml github/gh-aw compiler (control.md import + Pi/Copilot engine templates) upstream-owned context Medium secrets-outside-env (4006), overprovisioned-secrets (862) — compiler-generated auth-fallback expressions in every workflow None — report to github/gh-aw
zizmor all 70 .lock.yml github/gh-aw compiler upstream-owned context Low undocumented-permissions (416), template-injection (46, Pi/Copilot CLI invocation steps), obfuscation (1058), adhoc-packages (26, compiler-chosen npm install -g @earendil-works/pi-coding-agent / @playwright/cli) None — report to github/gh-aw
zizmor several .lock.yml (e.g. dashboard-authoring-corpus, design-decision-gate) github/gh-aw compiler (safe-outputs checkout template) upstream-owned context Medium artipacked — generated "Checkout repository (gh-aw default)" step after artifact download None — report to github/gh-aw
runner-guard self-care-hosted-health.lock.yml githubnext/gh-aw-cao (self-care-hosted-health.md) checking the already-authorized hosted CAO service target-owned, but not actionable as a vulnerability high (RGS-012) Flags the documented, intentional health-check curl to (cao.githubnext.com/redacted) (an explicitly declared, trusted hosted CAO endpoint) carrying a masked OIDC/read token for a read-only smoke test None — expected behavior for this self-care workflow; no change needed
runner-guard self-care-hosted-health.lock.yml github/gh-aw compiler (MCP Gateway template) upstream-owned context medium (RGS-019) Generated "Start MCP Gateway" step interpolates a step output in a run: block None — report to github/gh-aw
runner-guard agentic_commands.yml github/gh-aw compiler (slash-command router codegen) upstream-owned context medium (RGS-005) Generated router file grants write permissions under issue_comment/pull_request_review_comment triggers None — report to github/gh-aw
grype alpine:latest, ghcr.io/github/gh-aw-firewall/agent:0.28.44, and 7 other compiler-selected images Image publishers / github/gh-aw-firewall / github/gh-aw-mcpg / github/github-mcp-server upstream-owned context Critical/High/Medium/Low (277 unique CVE/GHSA IDs; highest: GO-2026-6612, GO-2026-6604 Critical in Go stdlib/golang.org/x/net) OS packages, Go stdlib, and npm (tar, http-cache-semantics) vulnerabilities in compiler-pinned runtime/firewall images None — report to owning image/component repository
grant same 9 images Image publishers upstream-owned context n/a (license policy) 178 license-policy findings (e.g. GPL-2.0-only busybox/alpine-baselayout, MPL-2.0 ca-certificates-bundle) against target .grant.yaml's permissive baseline None — target's .grant.yaml policy already exists; any policy change is a separate reviewed decision, not an image fix
Agent prompt
  1. Assign this issue to Copilot.
  2. Configure its MCP client to launch gh aw mcp-server over stdio from the target repository, then give it the prompt below. Require the server's fix and compile tools; never allow direct edits to generated .lock.yml files.
  3. Review the resulting pull request and rerun the same full compiler and security scan before merge. Require the target-owned actionable findings to be resolved; separately retained upstream-owned or unavailable evidence does not authorize target changes. If a target-owned finding needs human action, require the agent to stop and explain it.

Agent prompt

Fix only the target-owned actionable gh-aw compiler and security findings identified in this issue. Change only .github/workflows/*.md sources and directly related target-owned files; never edit generated .lock.yml files. Use the gh-aw MCP server's fix and compile tools, rerunning compilation with strict validation, model checks, actionlint, shellcheck, yamllint, zizmor, poutine, runner-guard, grant, grype, and syft until the target-owned findings are resolved. Review generated lock-file diffs and preserve existing behavior. Do not add or change .grant.yaml unless a separately reviewed target license-policy decision already requires it. Do not rebuild, modify, or make dependency decisions for upstream images or components; leave upstream-owned context unchanged and stop with a concise ownership explanation if the remaining finding cannot be fixed safely in this repository.

Specifically: in .github/workflows/shared/control.md, locate the cao_target_read_scope step (and any other steps matching the same pattern) whose run: block uses three sequential echo "..." >> "$GITHUB_OUTPUT" lines, and combine them into a single grouped redirect, e.g. { echo "owner=$owner"; echo "repository=$repository"; echo "full_name=$owner/$repository"; } >> "$GITHUB_OUTPUT", to resolve shellcheck SC2129 across all 70 compiled workflows.

Raw evidence
  • Compiler summary: Checked 70 workflow(s), Found 486 issue(s) (486 error(s)) — 139 shellcheck, 347 expression (all daily_ai_credits_*/safe-outputs-app-token template-wiring gaps).
  • zizmor: secrets-outside-env ×4006, obfuscation ×1058, overprovisioned-secrets ×862, undocumented-permissions ×416, artipacked ×49, template-injection ×46, adhoc-packages ×26 — all Low/Medium, spanning all 70 generated lock files identically (compiler-template origin confirmed: patterns are absent from every checked .md source and only appear in generated .lock.yml/agentic_commands.yml).
  • poutine: no findings (clean).
  • runner-guard: 3 findings total — RGS-012 (self-care-hosted-health, expected hosted-CAO health check), RGS-019 (self-care-hosted-health MCP Gateway template), RGS-005 (agentic_commands.yml router template).
  • syft: scanned 9 compiler-selected container images (alpine:latest; gh-aw-firewall agent/api-proxy/squid/cli-proxy 0.28.44; gh-aw-mcpg v0.4.29; gh-aw-node; github-mcp-server v1.12.2; node:lts-alpine).
  • grype: 277 unique CVE/advisory IDs across those 9 images (19 Critical, 177 High, 515 Medium-weighted rows, 119 Low).
  • grant: 178 license-policy violations, entirely within the same 9 upstream images, evaluated against the target's existing .grant.yaml.
  • No token-like or credential-like values were present in the reviewed tool output.
Control plane context

Generated by 🛡️ CAO Evolution / AW Compiler Security · copilot · auto · 271.6 AIC · ⌖ 12.6 AIC · ⊞ 16.2K · ◷

  • expires on Oct 26, 2026, 12:10 AM UTC
主要言語
JavaScript
スター
8
フォーク
5
平均マージ
2時間 50分
マージ済み PR（30日）
814

環境構築

はじめの一歩

  1. issue を最後まで読み、次にプロジェクトのコントリビューションガイドを読みます。
  2. 着手することを issue にコメントします — 二人が同じ作業をするのを防げます。
  3. リポジトリをフォークし、ブランチを切って変更します。
  4. issue 番号を参照したプルリクエストを送ります。

githubnext/gh-cao のほかの issue

githubnext/gh-cao の issue をすべて見る

似ている issue

JavaScript の issue をもっと見る

新しい issue をメールで受け取る

初心者向けの GitHub issue を短くまとめたダイジェスト。