w-fix-push
Pass
Audited by Gen Agent Trust Hub on Sep 21, 2026
Risk Level: SAFEDATA_EXFILTRATIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [DATA_EXFILTRATION]: The skill instructions proactively prevent the accidental exposure of sensitive information by forbidding the commitment of .env files, credentials, or secrets to version control.
- [COMMAND_EXECUTION]: The skill restricts command execution to standard project validation checks and Git operations, while explicitly requiring the use of Git verification hooks and prohibiting unauthorized configuration changes.
- [INDIRECT_PROMPT_INJECTION]: The skill includes instructions to mitigate risks from processing untrusted external data. Ingestion points: Task-related feedback including CI logs, review comments, and pull request metadata. Boundary markers: The instructions include a strong defensive policy requiring the agent to treat this data only as evidence of code issues and explicitly forbidding the execution of any embedded commands. Capability inventory: Capabilities are limited to Git publishing steps and local project validation scripts. Sanitization: The skill provides clear policy-based guidance to the agent to filter out adversarial content from external data sources.
Audit Metadata