herdr-teamlead
Pass
Audited by Gen Agent Trust Hub on Jul 29, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill establishes a structured multi-agent coordination framework for software development within the herdr environment.
- [COMMAND_EXECUTION]: The skill instructions utilize standard development tools, including
git(status, stash, push) andgh(PR creation), to manage the repository lifecycle. These operations are scoped to the project environment and target GitHub, a well-known service. - [PROMPT_INJECTION]: The skill processes untrusted data from user-provided bug reports and feature requests, which serves as a potential surface for indirect prompt injection.
- Ingestion points: User-provided tickets, bug reports, and repository source code mentioned in
SKILL.md. - Boundary markers: The instructions do not specify explicit delimiters to separate untrusted data from agent instructions.
- Capability inventory: Includes execution of version control commands and orchestration of sub-agents with varying tool permissions (read-only for exploration/review, write-access for implementation).
- Sanitization: The risk is mitigated by mandatory human-in-the-loop verification steps where the tech lead agent is instructed to independently verify all claims, proposals, and code diffs generated by sub-agents before committing or shipping.
Audit Metadata