The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes local shell commands via the subprocess module to integrate with Git for source fingerprinting and drift detection.
Evidence: assets/okf.py executes git rev-parse, git status, and git diff to identify changes in codebases being explained.
Mitigation: Commands are executed using argument lists rather than shell strings, preventing shell injection vulnerabilities.
[DYNAMIC_EXECUTION]: The skill instructions direct the agent to execute its own bundled Python scripts to manage knowledge persistence and scheduling.
Evidence: SKILL.md defines workflows involving python3 assets/okf.py and python3 assets/spaced_schedule.py.
Analysis: These scripts are part of the skill's core functionality, are written using the Python standard library only, and do not download or execute remote code.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an indirect prompt injection attack surface because its primary function is to ingest and process external content such as code repositories and research papers.
Ingestion points: Content is read from local files, Git repositories, and provided documentation via the agent's file system and network tools.
Boundary markers: The instructions do not explicitly mandate unique delimiters for source content, though they frame the interaction as an analysis and explanation task.
Capability inventory: The skill can write files (OKF bundles) and execute restricted local commands (git).
Sanitization: The assets/okf.py script includes a validation layer (validate_concept_meta) that prevents malformed or malicious metadata from corrupting the persistence format.

feynman-walkthrough