spec-compliance-auditor
Pass
Audited by Gen Agent Trust Hub on Oct 17, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the workspace that can influence its behavior and trigger command execution.
- Ingestion points: The skill reads task definitions and verification instructions from
specs/TASKS.mdand related specification files. - Boundary markers: Absent. The agent is instructed to read specific fields like
Verify:and execute them directly without any delimiters or warnings to ignore embedded instructions. - Capability inventory: The skill has the ability to execute arbitrary shell commands, perform git operations (commits), and modify files in the workspace.
- Sanitization: Absent. There is no logic to validate, escape, or filter the content of the commands retrieved from the task files before they are passed to the shell.
- [COMMAND_EXECUTION]: The skill instructions explicitly direct the agent to execute shell commands defined in workspace files.
- Evidence: The execution rules state: 'Execute the task's Verify: command yourself to confirm tests, typechecking, and linting genuinely pass.'
Audit Metadata