The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from Instagram, such as captions, comments, and profile information, which may contain malicious instructions designed to influence the agent.
Ingestion points: Data is retrieved from Instagram posts, reels, and user profiles using instagram-cli, hatch-zeitgeist, and browser tasks (referenced in SKILL.md and references/content-questions.md).
Boundary markers: The skill provides instructions to distinguish between "guesses" and "verified facts" and to report what could not be verified, although no explicit structural data delimiters are defined.
Capability inventory: The skill can execute commands via exec (including instagram-cli, hatch-zeitgeist, and ffmpeg), perform network requests to Meta/Instagram domains, and write files to the ~/workspace/ directory.
Sanitization: The agent is instructed to manually verify machine-generated annotations and is strictly prohibited from performing write actions, such as posting content or profile updates, without explicit user consent.
[COMMAND_EXECUTION]: The skill uses the exec command to run various CLI tools to perform its tasks.
Evidence: Instructions in SKILL.md and references/content-questions.md direct the agent to run instagram-cli, hatch-zeitgeist, and ffmpeg with arguments derived from user input, such as Instagram URLs and media IDs. The instruction templates include quoting for variables (e.g., '<exact supplied URL>') which helps mitigate common command injection risks.

instagram