improve-pipeline
Pass
Audited by Gen Agent Trust Hub on May 16, 2026
Risk Level: SAFEDATA_EXFILTRATIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [DATA_EXFILTRATION]: The skill is designed to collect context from the user's current workspace—including code examples, PR descriptions, and issue logs—and transmit this information to the vendor's central repository (chrislacey89/skills) by creating GitHub issues. This behavior is documented as the core purpose of the skill for pipeline evolution.
- [COMMAND_EXECUTION]: The workflow utilizes the GitHub CLI (gh) to view repository metadata, list issues, and create new issues or labels. It also executes a local /library command to query documentation stored on the user's filesystem.
- [PROMPT_INJECTION]: The skill ingests untrusted data from the project environment to build incident reports, creating an indirect prompt injection surface.
- Ingestion points: Local project artifacts (files, PRs, issues, logs) captured in Phase 1.
- Boundary markers: No explicit delimiters are used during data collection, although Phase 4 employs role-based agents (Advocate/Skeptic) to process the content.
- Capability inventory: GitHub issue/label creation via the gh CLI and local file modifications within the chrislacey89/skills repository.
- Sanitization: No specific sanitization or filtering of ingested incident data is described in the process.
Audit Metadata