tk-audit
Pass
Audited by Gen Agent Trust Hub on Sep 7, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONNO_CODE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill analyzes untrusted external data including PR reviews, issues, and CI logs. It includes a specific 'Retrieved Evidence Boundary' to treat such data as evidence rather than instructions.
- Ingestion points:
SKILL.mddefines the ingestion of external context from logs, web content, and transcripts. - Boundary markers: Explicit instructions in
SKILL.mdcommand the agent to ignore instruction-like text within evidence. - Capability inventory: The skill is limited to read-only analysis and local file logging, with no implementation or remote publication authority.
- Sanitization: Multiple files (
SKILL.md,audit-playbook.md,executor-handoff.md) prohibit the replication of secret values or credentials.
Audit Metadata