The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted external data which could contain malicious instructions targeting the agent.
Ingestion points: Data enters the agent's context through nlm source add --url and nlm research import as documented in SKILL.md and references/command_reference.md.
Boundary markers: The instructions do not define delimiters or specific "ignore" warnings to prevent the agent from following instructions embedded within the ingested sources.
Capability inventory: The agent utilizes bash for CLI operations and nlm download for file system writes, providing a mechanism for malicious instructions to trigger local actions.
Sanitization: There is no mention of sanitizing or filtering the text content retrieved from external sources before it is used in nlm notebook query or content generation tasks.
[COMMAND_EXECUTION]: The skill instructions explicitly direct the agent to use shell access for interacting with the nlm CLI.
Evidence: SKILL.md contains logic for detecting and using bash("nlm ...") commands to manage notebooks and sources.
[EXTERNAL_DOWNLOADS]: The skill performs network requests to fetch data from arbitrary external sources.
Evidence: The nlm source add and nlm research start commands fetch content from web URLs and YouTube to populate notebooks.

nlm-skill