qa-only
Pass
Audited by Gen Agent Trust Hub on Nov 16, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill relies on the Bash tool and a suite of local vendor binaries (e.g., gstack-skill-start, gstack-qa-evidence, gstack-qa-deadline) to perform testing probes, manage execution deadlines, and record evidence.
- [INDIRECT_PROMPT_INJECTION]: The skill has an inherent attack surface because its primary function is to ingest and report on data from untrusted sources like web pages, API responses, and CLI outputs. * Ingestion points: Browser DOM content, API responses, and process output captured during exploratory testing. * Boundary markers: The skill uses structured ID markers like gstack-qid:{question_id} and defined JSON schemas for exploration notes to separate data from instructions. * Capability inventory: The agent has access to Bash, Write, and Read tools, allowing for significant system interaction and file modification. * Sanitization: The instructions mandate the use of EVIDENCE_TOOL (gstack-qa-evidence), which is described as having built-in screening for credentials and sensitive data.
- [DYNAMIC_EXECUTION]: The gstack-skill-start command can inject dynamic instruction blocks (GSTACK_INSTRUCTION_BEGIN) into the agent's context. The skill mitigates this risk by requiring the agent to verify that these blocks carry a valid SESSION_ID and originate directly from the tool's output.
Audit Metadata