skill-authoring
Pass
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructions guide the agent to perform local environment setup and execute a validation script. This includes installing the
PyYAMLdependency and running thescripts/skill_lint.pytool to check for structural and descriptive quality in skills. Evidence:python -m pip install -r requirements.txtandpython scripts/skill_lint.pyinSKILL.md.\n- [INDIRECT_PROMPT_INJECTION]: The auditing workflow involves reading and analyzing external skill files, which represents an ingestion surface for untrusted data that could attempt to influence the agent's behavior during the audit. Audit instructions targetSKILL.mdand reference files.\n - Ingestion points: Reads target skill files (
SKILL.md,references/).\n - Capability inventory: Shell command execution (
pip,python) and file system access.\n - Boundary markers: The prompt does not define explicit delimiters to isolate content read from audited skills.\n
- Sanitization: The included linter script uses
yaml.safe_load()for frontmatter parsing, mitigating standard YAML-based injection attacks.
Audit Metadata