The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses a Python script, tokenop.py, which the agent is instructed to run locally to process large text files and command outputs.- [CREDENTIALS_UNSAFE]: The tokenop.py script includes a regex-based redaction mechanism that identifies common sensitive keywords like api_key, token, and password, replacing their values with <redacted> before the information is returned to the agent's context. It also maintains a local, unredacted archive of original text in a hidden project directory to allow for full evidence recovery.- [PROMPT_INJECTION]: The skill processes untrusted external data such as logs and tool results. While it includes logic to preserve important lines, it lacks explicit boundary markers or sanitization in the generated summaries, creating a potential surface for indirect prompt injection where the AI might misinterpret text in logs as instructions.

token-budget