diagnose-chromatic-baselines
Pass
Audited by Gen Agent Trust Hub on Oct 16, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill is a collection of diagnostic instructions and a local Python script for analyzing Chromatic visual testing issues. It does not perform any unauthorized network operations or download remote code.
- [DATA_EXFILTRATION]: The instructions contain extensive safety boundaries, explicitly prohibiting the agent from requesting API keys, tokens, session cookies, or project secrets. It includes a mandatory redaction policy for customer and committer PII (names, emails) and provides a
--redact-branchflag in the analysis script to protect sensitive branch names. - [COMMAND_EXECUTION]: The skill uses standard command-line utilities (
git merge-baseandgh pr view) to verify repository state. The arguments for these commands (commit SHAs and repository names) are constrained by the instructions and validated by the provided Python script'sfull_shatype-checker, reducing the risk of command injection. - [INDIRECT_PROMPT_INJECTION]: The skill has a surface for indirect prompt injection via the processing of external Chromatic CLI logs.
- Ingestion points: The
scripts/analyze_chromatic_log.pyscript reads and parses log files provided by the user. - Boundary markers: The
SKILL.mdfile defines a clear 'Evidence boundary' and provides a classification table to ensure the agent distinguishes between confirmed evidence and unverified data. - Capability inventory: The skill uses
python3to run a local script,gitfor ancestry checks, andghfor GitHub metadata. - Sanitization: The Python script uses strict regex patterns (e.g., 40-character hexadecimal SHAs) to extract data from logs, which prevents arbitrary log content from being misinterpreted as valid data fields by the agent.
Audit Metadata