maintain-verification-skill
Pass
Audited by Gen Agent Trust Hub on Aug 26, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes local verification harness scripts and launch commands as defined within the target verification skill's directory. This execution is scoped to project-local files and is a primary function of the maintenance process.
- [PROMPT_INJECTION]: The skill implements an indirect prompt injection surface by using subagents to read and summarize project source code files. This is a standard architectural pattern for code-analysis tools.
- Ingestion points: The skill reads feature files from the project's source code during the 'Source wave' phase to identify documentation drift.
- Boundary markers: While it instructs subagents to produce concise summaries and recipes, there are no specific markers or delimiters defined in the instructions to isolate the source code content from the agent's instructions.
- Capability inventory: The skill possesses the ability to modify local documentation files (.md), update harness scripts, and execute shell commands for application verification.
- Sanitization: No explicit sanitization or filtering of the summaries generated from source code is specified.
Audit Metadata