The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands that pipe data from a remote URL to a local Node.js script for validation: curl -fsSL https://{domain}/.well-known/releases.json | node <skill-dir>/scripts/validate.mjs /dev/stdin. While the local script performs data validation rather than code execution, this pattern involves processing external input through shell pipelines.
[EXTERNAL_DOWNLOADS]: The skill fetches content from user-specified external domains and interacts with a remote API at api.releases.sh to validate listings. These operations are core to the skill's purpose but involve outbound network requests to resources outside the immediate control of the skill.
[INDIRECT_PROMPT_INJECTION]: The skill processes releases.json manifest files which are hosted on external, untrusted domains, creating a potential vector for indirect prompt injection.
Ingestion points: The curl command in SKILL.md retrieves manifest data from a user-provided {domain}.
Boundary markers: There are no explicit instructions or delimiters used when fetching the remote data to prevent the agent from following instructions embedded within the manifest content.
Capability inventory: The skill utilizes curl for network access, node for running a local validation script, and has the ability to write JSON files to the local repository.
Sanitization: The scripts/validate.mjs script provides schema validation and type checking for the JSON content, though the agent may still be exposed to malicious text strings if they are reflected in validation error messages or processed during the manifest creation phase.

creating-releases-json