handoff
Pass
Audited by Gen Agent Trust Hub on Jul 20, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection when resuming context from existing handoff documents. An attacker could craft a malicious handoff file that, when loaded by an agent via the
--resumeflag, contains instructions intended to manipulate the agent's behavior or trigger unintended skill invocations. - Ingestion points: The skill reads external markdown files from the filesystem during Workflow B (Resume).
- Boundary markers: Absent. The skill does not implement delimiters or explicit 'ignore instructions' warnings for content parsed from the handoff documents.
- Capability inventory: The skill is capable of listing and initiating the invocation of other installed skills (e.g.,
commit,make-pr,peer-review) based on the contents of the## Suggested skillssection in the resumed file. - Sanitization: Absent. While the skill performs redaction when creating new documents, there is no evidence of content sanitization or instruction filtering when loading existing files.
Audit Metadata