diagnosing-bugs
Pass
Audited by Gen Agent Trust Hub on Nov 7, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection because it is designed to process untrusted external data such as bug reports, logs, and captured network traces (HAR files).
- Ingestion points: SKILL.md (Phase 1) describes ingesting HAR files, log dumps, and core dumps provided by users.
- Boundary markers: No specific delimiters for untrusted data are provided.
- Capability inventory: The skill allows the agent to execute shell commands (curl, git, test runners), read/write files, and modify code.
- Sanitization: The skill mandates the redaction of secrets and authentication headers (using ) before any data is processed or shown to the user.
- [DYNAMIC_EXECUTION]: The skill instructs the agent to generate and execute temporary scripts, reproduction harnesses, and tests as part of the debugging loop. This is a standard and necessary function of a bug diagnosis tool.
- [COMMAND_EXECUTION]: The agent is authorized to run common development command-line tools to validate hypotheses and reproduce failures.
Audit Metadata