The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/jd_classify.py uses importlib.util to dynamically load and execute functions from a computed relative path (../../diataxis/scripts/diataxis_classify.py). This dynamic execution pattern is used for optional Diataxis framework integration.\n- [REMOTE_CODE_EXECUTION]: Documentation in SKILL.md and TROUBLESHOOTING.md recommends installing a dependency by piping a remote script to a shell: curl -LsSf https://astral.sh/uv/install.sh | sh. While the source is a well-known service, this method involves executing remote code without local verification.\n- [COMMAND_EXECUTION]: The utility script scripts/shared.py executes the git command using subprocess.run to determine the project root directory, which is a common operation for project-aware tooling.\n- [PROMPT_INJECTION]: The skill presents an indirect prompt injection surface when processing project documentation. Ingestion points: The scripts jd_classify.py and jd_index.py read markdown file headings and content from the user's project. Boundary markers: Absent; the skill does not use delimiters to isolate untrusted file content. Capability inventory: Subprocess calls in shared.py, file move operations in jd_add.py and jd_classify.py, and file write operations in jd_init.py, jd_add_area.py, and jd_index.py. Sanitization: Absent; while filenames are normalized and content is lowercased for keyword matching, there is no filtering of embedded natural language instructions.

jd-docs