The Agent Skills Directory

[SAFE]: The skill consists of markdown instructions and template structures for a punishment-based accountability system. It does not contain any executable scripts, binary files, or remote code dependencies.
[DATA_EXPOSURE]: The skill documents the use of a hidden directory in the user's home folder (~/.openanalyst/) for persistent storage of contract details, including contact information like names and email addresses of referees.
[INDIRECT_PROMPT_INJECTION]: The skill's workflow depends on external input from a 'referee' to verify commitment completion, which presents a surface for potential indirect prompt injection.
Ingestion points: The agent reads state from ~/.openanalyst/contracts/active-contract.json and processes confirmation messages or appeals provided by human referees.
Boundary markers: The instructions do not include specific delimiters or 'ignore' instructions to separate referee-provided content from the agent's core instructions.
Capability inventory: The skill design implies the ability to perform file system operations (read/write) for contract management and potentially interact with messaging systems to notify referees.
Sanitization: There are no documented procedures for sanitizing or validating input received from external referees before it is integrated into the agent's decision-making context.

punishment