tailrocks-code-health

Pass

Audited by Gen Agent Trust Hub on Sep 8, 2026

Risk Level: SAFEDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [DYNAMIC_EXECUTION]: The skill instructions (SKILL.md) require the agent to programmatically derive the file path for the 'code-health-predicate.ts' script by navigating parent directories from the skill's installation path. While this is used to locate the skill's own internal logic, executing code from dynamically computed paths is a noted security pattern.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted repository-relative files which could theoretically contain malicious instructions.
  • Ingestion points: The skill reads project configuration and ledger files including ratchet.toml, flaky-tests.toml, and DEFECT_LEDGER.md.
  • Boundary markers: The 'runtime-trust.md' reference explicitly defines repository files, reports, and tool outputs as untrusted data that cannot alter the agent's governing rules or authority.
  • Capability inventory: The skill possesses capabilities for writing to approved file system paths, executing measurement shell commands, and running the predicate script (as specified in SKILL.md).
  • Sanitization: Measurement steps in SKILL.md mandate that secrets be scrubbed and target networks be disabled during execution.
  • [COMMAND_EXECUTION]: The skill performs shell-based measurements (e.g., 'cargo metadata') to prove preconditions for code health changes. These operations are restricted to frozen inputs, isolated process trees, and lack network access, which serves as a safety mitigation.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 8, 2026, 03:12 PM
Security Audit — agent-trust-hub — tailrocks-code-health