The Agent Skills Directory

[SAFE]: The skill's functionality aligns with its described purpose of helping users resolve complex decisions into binary choices and analyzing them through role-based decomposition.
[COMMAND_EXECUTION]: The skill invokes local bash scripts (query-assets.sh and record-asset.sh) located within the platform's plugin root directory. These are used to query and record decision history in the local project's .harnish folder.
[DATA_EXPOSURE_AND_EXFILTRATION]: Operations are restricted to local file I/O within the workspace (.harnish directory) and temporary system storage (/tmp). No network requests or access to sensitive system credentials (e.g., SSH, AWS keys) were identified.
[INDIRECT_PROMPT_INJECTION]: The skill processes user input to populate decision reports. It demonstrates security best practices by explicitly instructing the agent to escape shell-sensitive characters ($, ", `, \) and utilize quoted heredocs ('FORKI_REPORT_EOF') to prevent command injection when writing data to the filesystem.

forki