debugging-strategies
Pass
Audited by Gen Agent Trust Hub on Oct 27, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to ingest and analyze external data such as logs, traces, and environment details. This surface is susceptible to indirect prompt injection if the processed data contains malicious instructions intended to manipulate the agent.
- Ingestion points: Instructions in SKILL.md to "capture logs, traces, and environment details" and the "Information Collection" phase in resources/implementation-playbook.md.
- Boundary markers: The skill lacks explicit instructions for the agent to use delimiters or to disregard natural language instructions found within logs and traces.
- Capability inventory: The skill references capabilities including shell command execution (git bisect, delve), debugger interaction, and file system operations (heap snapshots).
- Sanitization: No sanitization or validation logic is provided for the ingested data.
- [EXTERNAL_DOWNLOADS]: The implementation playbook suggests installing an external debugging tool from a well-known source.
- Evidence: The command 'go install github.com/go-delve/delve/cmd/dlv@latest' is referenced in resources/implementation-playbook.md. The download targets a known official repository for the Delve debugger.
Audit Metadata