The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes markdown files from the user-controlled directory ~/.claude/style-library/. These files contain style anchors that are directly injected into prompts for other tools. While the skill performs schema validation, it does not sanitize or filter the natural language content, allowing for potential instructions to be embedded in style definitions. 1. Ingestion points: Style files located in ~/.claude/style-library/. 2. Boundary markers: Absent; files are processed as raw markdown. 3. Capability inventory: Includes Bash for CLI execution and Write/Edit for file management. 4. Sanitization: Only schema-level validation is performed; no content-based filtering of style anchors.
[DYNAMIC_EXECUTION]: The entry script scripts/run.py implements process replacement using os.execv() to ensure the logic runs within a specific virtual environment. While this is a standard bootstrapping pattern, it involves replacing the active process with a new one based on paths computed at runtime. Evidence: os.execv(str(_VENV_PY), [str(_VENV_PY), str(HERE), *sys.argv[1:]]).
[COMMAND_EXECUTION]: The skill's operation relies on the execution of a Python-based CLI via the Bash tool, as specified in the allowed-tools and PIPELINE sections of SKILL.md. This allows for broad interaction with the filesystem and external runners within the scope of the provided scripts.

skills-styles