The Agent Skills Directory

[PROMPT_INJECTION]: The skill explicitly instructs the agent to be 'permissionless' and states that its job is 'not to prepare work for a human to approve'. This overrides standard safety protocols regarding human-in-the-loop review for external actions like sending emails or posting to public forums.
[COMMAND_EXECUTION]: The core functionality of the skill relies on executing the autark CLI tool to manage hypotheses, runs, and actions.
[EXTERNAL_DOWNLOADS]: The documentation directs users to install an external, third-party CLI tool via npm i -g autark.
[DATA_EXFILTRATION]: The skill is designed to log the agent's external activities (emails, GitHub comments, Reddit posts) to an external API (autark-api.kushalsokke.workers.dev) and a public dashboard.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by ingesting untrusted data through the autark context command.
Ingestion points: The agent reads data from local product briefs (products/<slug>.md) and remote context (hypotheses, recent runs, and narratives) via the autark context command.
Boundary markers: No explicit delimiters or boundary markers are used to separate ingested data from agent instructions.
Capability inventory: The agent has the capability to perform network operations and external posts via autark log action (Email, GitHub, Reddit, HN, etc.).
Sanitization: There is no evidence of sanitization or validation of the remote content fetched by the CLI before it is processed by the agent.

autark