The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the Bash tool to dynamically derive file paths for state persistence. It executes shell commands (git rev-parse, PWD, tr) to generate a unique 'slug' for the session's storage directory.
[PROMPT_INJECTION]: The skill instructs the agent to auto-invoke the pinning mechanism without user permission whenever specific triggers (decisions, questions, or constraints) are detected in conversation. This 'do not ask permission' instruction reduces user oversight of automated tool execution.
[PROMPT_INJECTION]: The skill establishes a surface for indirect prompt injection by persisting untrusted conversation data and re-injecting it into the model's future tool-calling context.
Ingestion points: User conversation inputs matching defined triggers (e.g., 'yes go with bun', 'what about latency?') as described in the 'Auto-Invoke Rules' section of SKILL.md.
Boundary markers: Absent. The skill explicitly commands the agent to 'Store verbatim. No cleanup, no categorization, no reformulation.'
Capability inventory: The skill possesses Bash, Read, Write, and Edit tools (specified in the allowed-tools frontmatter), which could be targets for injected instructions.
Sanitization: Absent. Content is written directly to the pins.json file without validation or escaping.

pin