harvest-run
Pass
Audited by Gen Agent Trust Hub on Aug 25, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to use git commands (
git rmandgit commit) to manage file deletions and commit changes to the repository. This is a core part of the skill's intended cleanup functionality. - [INDIRECT_PROMPT_INJECTION]: The skill processes content from potentially untrusted files within a run directory and an external handoff document, creating a surface for indirect prompt injection.
- Ingestion points: Reads every file in the specified run directory and the handoff document as defined in
SKILL.md. - Boundary markers: None identified; there are no instructions to the agent to ignore or delimit instructions found within the ingested files.
- Capability inventory: The skill has the ability to write to files (decisions, lessons, and backlog in the
docs/folder) and delete directories usinggit rm. - Sanitization: The skill does not specify any sanitization, filtering, or validation of the content extracted from the run files before it is committed to the repository's durable documentation.
Audit Metadata