The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes the threads-cli tool to interact with the Threads API, providing a controlled environment for account management tasks as specified in SKILL.md.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from the Threads platform through search, feed, and post comments (SKILL.md). Mandatory Evidence: (1) Ingestion points: search, feed, post, and comments commands; (2) Boundary markers: uses the social_posts_v1 compact presentation format for results; (3) Capability inventory: shell command execution via the CLI and publish-post write capabilities (manifest.yaml); (4) Sanitization: relies on provider-side filtering and server-side invariants for data safety.
[DATA_EXFILTRATION]: The skill retrieves sensitive personal data including activity notifications, liked media, and private account insights. This access is authorized via OAuth (connector_auth_config.yaml) and requires explicit account linking through the Meta Accounts Center.
[PRIVILEGE_ESCALATION]: The publish-post command allows reading files from the /workspace/ directory for media uploads. The instructions strictly prohibit access to temporary or system-level paths like /tmp, ensuring operations remain within the defined agent workspace boundaries.

threads