The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to search and play back record files, which introduces a surface for processing untrusted data.
Ingestion points: The skill reads JSONL record files from the ${CLAUDE_PLUGIN_DATA}/records/ directory and a user-configurable mirror folder (SKILL.md).
Boundary markers: There are no explicit instructions or delimiters provided to the agent to treat content within these records as data only and to ignore any embedded instructions.
Capability inventory: The agent has the capability to list directories, search file text, read contents, and append new lines to these records (SKILL.md).
Sanitization: No sanitization or validation of the text fields within the records is mentioned before they are displayed using the playback template.
[COMMAND_EXECUTION]: The skill instructs the agent to perform directory listing and file searching, and explicitly mentions using a shell command for timestamping.
Evidence: The instruction in SKILL.md suggests using date -u +%Y%m%dT%H%M%SZ to generate filenames, which involves executing a system utility.
[EXTERNAL_DOWNLOADS]: The documentation provides instructions for installing the skill and its dependencies from a remote repository.
Evidence: The README.md includes an installation command npx skills add https://github.com/synapseradio/ai-skills. This refers to the author's own repository and is a standard setup procedure for this ecosystem.

check-notes