The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill suggests installing napkin-ai globally via NPM and automates the installation of the pi-napkin package at a project-local scope using the pi tool.
[COMMAND_EXECUTION]: The skill utilizes several CLI commands to manage the Napkin vault and Pi environment, including napkin init, napkin config set, and pi install. It specifically uses the --approve flag in the pi tool, which may bypass manual security confirmations during execution.
[DATA_EXPOSURE]: The skill reads sensitive configuration files from the user's home directory (~/.pi/agent/settings.json) and the project root (.pi/settings.json) to determine the current extension state and model routing policies.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from repository files like NAPKIN.md and docs/agents/*.md to generate summaries and instruction pointers.
Ingestion points: Files processed include NAPKIN.md, AGENTS.md, CLAUDE.md, and markdown files within docs/agents/ (e.g., issue trackers, domain glossaries).
Boundary markers: The skill does not define explicit boundary markers or use "ignore embedded instructions" warnings when processing these files.
Capability inventory: The skill can execute shell commands (napkin, pi), write to the filesystem, and modify project configuration files (.pi/settings.json).
Sanitization: There is no evidence of sanitization or filtering of content ingested from the vault files before it is used to draft instructions or summaries.

integrate-napkin-agent-skills