The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from external sources such as git commit history and user interviews, which could contain malicious instructions designed to influence the agent's behavior during the drafting process.
Ingestion points: Processes repository metadata and commit bodies using git log and git show --stat as described in SKILL.md, and incorporates user feedback through interactive interviews (/grill-me).
Boundary markers: Although the skill uses structured Markdown and YAML frontmatter, it lacks specific delimiters or guardrail instructions to ensure the agent ignores potentially adversarial content within the ingested git data.
Capability inventory: The agent is authorized to write files (CANONICAL.md, platform copy files), execute shell commands for the Buffer CLI, and interact with system clipboard utilities.
Sanitization: No explicit sanitization, filtering, or validation routines are specified for data retrieved from git repositories or user inputs.
[COMMAND_EXECUTION]: The workflow includes instructions for the agent to execute shell commands to facilitate social media publishing and local utility interactions.
Evidence: SKILL.md and references/linkedin_playbook.md provide command patterns for the buffer CLI (buffer posts create) and system clipboard tools including pbcopy (macOS), xclip (Linux/X11), and wl-copy (Linux/Wayland).

social-copy