cli-spec
Pass
Audited by Gen Agent Trust Hub on Apr 12, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTIONDATA_EXFILTRATION
Full Analysis
- [PROMPT_INJECTION]: The 'references/evals.yaml' file contains several strings that match prompt injection patterns (e.g., 'JAILBREAK', 'Ignore all previous instructions', 'Override all safety policies'). These are explicitly used as test cases to evaluate the skill's security robustness and do not function as instructions for the agent's actual behavior.
- [COMMAND_EXECUTION]: The 'SKILL.md' file instructs the agent to run diagnostic scripts located in the local repository ('python3 scripts/diagnose_skill.py' and 'python3 utilities/skill-builder/scripts/quick_validate.py') to validate generated artifacts.
- [DATA_EXFILTRATION]: The skill design incorporates strict redaction policies, as seen in 'SKILL.md' and 'references/extended.md', which require the removal of secrets, API keys, and PII from all sample outputs and logs.
- [PROMPT_INJECTION]: This skill possesses an indirect prompt injection surface because it processes untrusted user input (e.g., brainstorms and feature descriptions) to create technical contracts. Ingestion points include the primary user request and existing repository files accessed during research. While the skill lacks explicit boundary markers to separate untrusted data from instructions, it includes capabilities for reading and writing files and emphasizes data sanitization (redaction).
Audit Metadata