Repository navigation
chore: materialize APM primitives - #233
devsecninja-automation[bot] wants to merge 1 commit into
Conversation
|
Important Review skippedBot user detected. To trigger a single review, invoke the
|
SkillSpector report:
|
| Severity | Reported | Accepted | Blocking |
|---|---|---|---|
| CRITICAL | 0 | 0 | 0 |
| HIGH | 17 | 17 | 0 |
| MEDIUM | 50 | - | - |
| LOW | 2 | - | - |
Top findings
| Severity | Rule | Location | Finding | Status |
|---|---|---|---|---|
| HIGH | AE1 | .apm/skills/obsidian/SKILL.md:217 | Incomplete referenced artifact analysis: references/FUNCTIONS_REFERENCE.md (partial) | Accepted: Derived from the accepted static_parse_limit on FUNCTIONS_REFERENCE.md: a Markdown link to bundled reference docs, not a risky artifact. |
| HIGH | AR2 | .apm/skills/renovate-triage-and-merge/references/testing.md:38 | Anti-Refusal Statement: no warning | Accepted: Test expectation that happy-path results emit no warning; not an instruction to suppress refusals. |
| HIGH | AR2 | .apm/skills/renovate-triage-and-merge/references/testing.md:78 | Anti-Refusal Statement: without warnings | Accepted: Test expectation that valid ports are accepted without warnings; not an instruction to suppress refusals. |
| HIGH | E2 | .apm/skills/renovate-triage-and-merge/scripts/renovate_prs.py:205 | Env Variable Harvesting: os.environ.copy() | Accepted: Passes the caller's environment to the local gh CLI so it can authenticate; nothing is read out or sent elsewhere. |
| HIGH | YR4 | .github/agents/project-architecture-planner.agent.md:3 | YARA rule 'agent_skill_mcp_tool_poisoning_metadata': MCP/tool metadata poisoning indicators in tool schemas or skill manifests [agent_skills]: tools:; description:; <!-- System boundary --> | Accepted: External awesome-copilot agent: the HTML comment labels a draw.io diagram example, not a hidden tool instruction. |
| HIGH | P2 | .github/agents/project-architecture-planner.agent.md:396 | Hidden Instructions: <!-- System boundary --> | Accepted: External awesome-copilot agent: XML comment inside a draw.io diagram example, not a hidden instruction. |
| HIGH | TM1 | .github/agents/sast-sca-security-analyzer.agent.md:319 | Tool Parameter Abuse: subprocess.call(cmd, shell=True | Accepted: External awesome-copilot agent: listed as a vulnerable pattern to detect (CWE-78), not code to run. |
| HIGH | TM1 | scripts/git-push-approval/require-push-approval.sh:3 | Tool Parameter Abuse: Git Push Approval Hook # Forces an explicit, interactive human approval before the AI coding agent runs # any `git push`. A commit is entirely local and rewritable (amend / reset / # rebase) until it | Accepted: Push-approval hook: it names --force options only to detect them and ask for approval; it never runs them. |
| HIGH | TM1 | scripts/tool-guardian/guard-tool.sh:86 | Tool Parameter Abuse: rm -rf / | Accepted: Tool Guardian's deny-list: these destructive commands are the regexes it blocks, never commands it runs. |
| HIGH | TM1 | scripts/tool-guardian/guard-tool.sh:86 | Tool Parameter Abuse: 'rm' on specific paths instead of root" "destructive_file_ops:::critical:::rm -rf ~:::Use targeted 'rm' on specific paths instead of home directory" "destructive_file_ops:::critical:::rm -rf \ | Accepted: Tool Guardian's deny-list: these destructive commands are the regexes it blocks, never commands it runs. |
Showing 10 of 69 findings, blocking first; see Code Scanning for all.
Errors
- .github/skillspector-accepted.json: Stale findings entry for .agents/skills/security-audit/AI-AND-LLM.md (MP3); remove it or update it to the current report.
- .github/skillspector-accepted.json: Stale findings entry for .agents/skills/security-audit/ATTACK-CLASSES.md (PE3); remove it or update it to the current report.
- .github/skillspector-accepted.json: Stale findings entry for .agents/skills/security-audit/DESKTOP-MOBILE-AND-LOCAL-IPC.md (PE3); remove it or update it to the current report.
- .github/skillspector-accepted.json: Stale findings entry for .agents/skills/security-audit/SKILL.md (AE1); remove it or update it to the current report.
- .github/skillspector-accepted.json: Stale findings entry for .agents/skills/security-audit/SKILL.md (AE1); remove it or update it to the current report.
- .github/skillspector-accepted.json: Stale findings entry for .agents/skills/security-audit/validate-coverage-ledger.test.cjs (PE3); remove it or update it to the current report.
- .github/skillspector-accepted.json: Stale findings entry for .agents/skills/security-audit/validate-findings.test.cjs (PE3); remove it or update it to the current report.
- .github/skillspector-accepted.json: Stale findings entry for .agents/skills/security-audit/validate-findings.test.cjs (PE3); remove it or update it to the current report.
- .github/skillspector-accepted.json: Stale limitations entry for .agents/skills/security-audit/SKILL.md (reference_missing); remove it or update it to the current report.
- .github/skillspector-accepted.json: Stale limitations entry for .agents/skills/security-audit/SKILL.md (reference_missing); remove it or update it to the current report.
Analysis limitations (16)
| Target | Location | Reason | Status |
|---|---|---|---|
| .apm/skills/commit-and-release | .apm/skills/commit-and-release/SKILL.md:171 | reference_missing | Caveat, non-blocking |
| .apm/skills/commit-and-release | .apm/skills/commit-and-release/SKILL.md:173 | reference_missing | Caveat, non-blocking |
| .apm/skills/commit-and-release | .apm/skills/commit-and-release/SKILL.md:181 | reference_missing | Caveat, non-blocking |
| .apm/skills/meeting-transcript-creation | .apm/skills/meeting-transcript-creation/SKILL.md:17 | reference_missing | Caveat, non-blocking |
| .apm/skills/morning-briefing | .apm/skills/morning-briefing/SKILL.md:49 | reference_missing | Caveat, non-blocking |
| .apm/skills/morning-briefing | .apm/skills/morning-briefing/SKILL.md:58 | reference_missing | Caveat, non-blocking |
| .apm/skills/obsidian | .apm/skills/obsidian/SKILL.md:32 | reference_missing | Accepted: The path-like reference is generic `[text](url)` link syntax in prose, not a bundled file. |
| .apm/skills/obsidian | .apm/skills/obsidian/references/FUNCTIONS_REFERENCE.md:69 | static_parse_limit | Accepted: Obsidian Bases formula examples in a reference doc; the parser cannot bound them as commands but checked all input. |
| .apm/skills/renovate-triage-and-merge | .apm/skills/renovate-triage-and-merge/SKILL.md:198 | reference_missing | Caveat, non-blocking |
| .apm/skills/renovate-triage-and-merge | .apm/skills/renovate-triage-and-merge/SKILL.md:204 | reference_missing | Caveat, non-blocking |
Showing 10 of 16 limitations; see the run summary for all.
Full results: Code Scanning alerts · workflow run and full reports
f5e2453 to
c839eb0
Compare
c839eb0 to
3b03e59
Compare
|
Closing unmerged: this PR would delete the committed It's an apm-cli 0.35.0 bug (microsoft/apm#3372), not a real upstream change. On a fresh runner with no per-user deployment receipts, APM skips every committed skill file ( #240 fixes APM Materialize: it removes the lock-recorded external skill folders before |
Automated PR — materializes the APM primitives for the current
apm.yml. Runsapm installto refreshapm.lock.yamland redeploythe external agents into
.github/agents/and skills into.agents/skills/after anapm.ymlchange landed onmain(e.g. aRenovate SHA/tag bump).
Review the
.github/agents/and.agents/skills/diff — thisis the prompt-hijacking gate: confirm the upstream text changed as
expected. Local pins from
apm-overrides.jsonare re-applied on top.Never auto-merge this PR: the owner reviews and merges it. See
docs/apm.md.Triggered by
APM Materializeat commitb2945ba6777172d9d2f2ee8da6aedae8525cbc5a.