The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute local utility and test scripts (python scripts/handoff_status.py and python tests/run_tests.py) to gather empirical data for its grading report. These operations are restricted to the local environment and serve the skill's primary purpose of system assessment.
[INDIRECT_PROMPT_INJECTION]: The skill exhibits an ingestion surface for untrusted data as it reads and processes multiple log files and wiki pages (e.g., qa-rejection-log.md, ann-overlay.md) which could theoretically contain malicious instructions.
Ingestion points: Multiple markdown log files and documentation in the agent-improvements/ and mel_wiki/ directories, as well as the agent registry.
Boundary markers: The skill instructions specify strict analytical workflows, directing the agent to "extract retrospective bullets," "count sessions," and "trend the verdicts," which treats the data as objects of analysis rather than a source of instructions.
Capability inventory: The skill is capable of reading local files and executing specific project scripts; it does not have network access or arbitrary file-write capabilities.
Sanitization: No explicit sanitization of the input data is performed, but the output is restricted to a structured report format defined in a template, which limits the impact of potentially injected content.

grade-system