The Agent Skills Directory

[DATA_EXFILTRATION]: The skill identifies and provides instructions for interacting with sensitive filesystem locations, such as ~/Documents/DobbySecrets and ~/.secrets. These paths are part of the intended environment configuration for secret management.
[COMMAND_EXECUTION]: The instructions direct the agent to utilize a local script, ~/GitHub/scripts/bin/local-secrets, to perform operations such as reading, writing, and validating secrets.
[INDIRECT_PROMPT_INJECTION]: The skill defines a surface for processing untrusted user data (secret names and values) that are interpolated into configuration files and shell commands. * Ingestion points: Secret data provided by the user during the management workflow in SKILL.md. * Boundary markers: No explicit delimiters or boundary markers are defined for the handling of user-provided secret values. * Capability inventory: The skill references file write operations for .env files and shell execution via a local script. * Sanitization: A specific guardrail is included instructing the agent to never print secret values in logs or responses.

secret-management