security-response-red-purple-team
Pass
Audited by Gen Agent Trust Hub on Sep 16, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted data from project-level evidence files to determine security coverage status.\n
- Ingestion points: Processes data from
.appsec/state.json,.appsec/evidence/<tag>/threat-model/, and various findings files (e.g.,.appsec/findings/<tag>/*.yaml).\n - Boundary markers: Relies on structured YAML templates for output but lacks explicit delimiters for untrusted input ingestion within the workflow steps.\n
- Capability inventory: Utilizes
Bash(for SDK interaction),Read,Grep, andGlobtools.\n - Sanitization: Implements a
redact_on_outputpolicy in the frontmatter to strip credentials and PII from generated reports.\n- [COMMAND_EXECUTION]: The skill uses theBashtool to perform operations through a local script.\n - Evidence: Invokes
bash ~/.claude/scripts/appsec-sdk.shto persist theattack-coverage.yamlartifact.\n - Context: While intended for legitimate SDK interactions, the use of a shell interpreter with dynamic arguments (such as the
<tag>variable) represents a standard command execution surface.
Audit Metadata