The Agent Skills Directory

[SAFE]: The skill implements a local documentation management system for Architecture Decision Records (ADRs). The core logic resides in scripts/decisions.py, which uses standard Python libraries to perform regex-based text processing and file management.
[COMMAND_EXECUTION]: The install command in scripts/decisions.py performs administrative setup tasks: it creates a machine-specific symlink for the script, updates the .gitignore file, and installs a git pre-commit hook. These actions are standard for developer tooling and are clearly documented in the skill's instructions.
[DATA_EXPOSURE]: The residual_mnemonics function performs a recursive scan of the project repository (excluding common directories like .git and node_modules) to identify instances where old mnemonics might still exist in code or prose. This is a read-only operation intended to provide a report to the user and does not involve network activity or data exfiltration.
[INDIRECT_PROMPT_INJECTION]: The script parses YAML-like front-matter and markdown content from ADR files. While it processes external data, the parsing is handled via controlled regex patterns and custom logic rather than unsafe evaluation (like eval() or yaml.load()). Boundary markers are maintained through the ADR template structure.

decision-records