The Agent Skills Directory

[DATA_EXFILTRATION]: The skill communicates with an external API service defined by the PYLOT_API environment variable. It transmits a PYLOT_DISPATCH_TOKEN (Bearer token) to this endpoint to authenticate requests for team metrics, costs, and mission data. While the domain is user-configurable, this represents a flow of sensitive authentication material to a network-connected service.
[COMMAND_EXECUTION]: The skill uses the Bash tool to execute a series of shell commands involving curl, jq, and the GitHub CLI (gh). It interpolates user-supplied arguments (specifically the team argument) into shell variables and subsequent command strings. While primarily intended for local workflow orchestration, these commands interact with external systems and the local file system to manage planning artifacts.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external sources, specifically GitHub issue descriptions, Pull Request comments, and mission history, which could contain instructions designed to influence the agent's behavior during the summary and recommendation phases.
Ingestion points: Data is ingested via gh issue list, gh pr list, and PYLOT_API/missions API calls (processed in SKILL.md Steps 0, 1, and 2).
Boundary markers: No explicit delimiters or boundary markers are defined in the instructions to isolate untrusted GitHub content from the agent's primary task instructions.
Capability inventory: The skill possesses significant capabilities, including executing shell commands (Bash), writing to the file system (Write), and making state-changing API calls (PUT, POST, PATCH) to update goals and budgets.
Sanitization: The skill relies on a mandatory 'human-in-the-loop' confirmation process (documented in the 'Boundaries' and 'Anti-patterns' sections) before any destructive or irreversible actions are taken, which serves as a primary mitigation against automated exploitation.

weekly-plan