The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a git pull command to synchronize local data with a remote repository. This operation is performed automatically upon locating the repository root marked by a .ceos file. While this is a functional requirement for the skill's purpose, it executes a shell command with a dynamically determined path.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection as it reads and parses organizational data from external files which could contain malicious instructions.
Ingestion points: Data is ingested from data/accountability.md, templates/accountability.md, and the data/people/ directory.
Boundary markers: The skill does not implement specific delimiters or 'ignore' instructions when parsing seat titles, owner names, or role descriptions from the markdown files.
Capability inventory: The agent has permissions to Read, Write, Edit, and Glob files, as well as execute git commands. It can modify the source of truth for organizational data.
Sanitization: There is no evidence of input validation or content filtering to prevent the agent from interpreting data as instructions. However, the skill explicitly mandates a human-in-the-loop 'diff' review and approval process before any write operation is performed.

ceos-accountability