code-autonomous

Pass

Audited by Gen Agent Trust Hub on Jul 24, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection by processing untrusted data to guide autonomous agent actions.\n
  • Ingestion points: Untrusted data enters the agent context via 'descriptors' (local plan files, pasted text, or GitLab issue payloads) as mentioned in SKILL.md and references/standalone-pipeline.md.\n
  • Boundary markers: The subagent prompt template in references/multi-agent-dispatch.md interpolates variables like {unit description} without surrounding them with delimiters or providing instructions to ignore embedded commands.\n
  • Capability inventory: The skill has the capability to perform file system writes (via the code-coder subagent), create git worktrees, and execute git commits.\n
  • Sanitization: No sanitization or validation of the input descriptor content is described in the provided files.\n- [COMMAND_EXECUTION]: The skill requires the execution of shell commands to establish its isolated work environment.\n
  • Evidence: references/standalone-pipeline.md specifies the use of 'git worktree add' to create an isolated workspace and 'git rev-parse' and 'pwd' for preflight verification before any code edits occur.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 24, 2026, 05:00 PM
Security Audit — agent-trust-hub — code-autonomous