The Agent Skills Directory

Indirect Prompt Injection Surface: The skill is designed to analyze external project documentation, including briefs, issue trackers, and commit history. The ingestion of untrusted data from these external sources serves as a potential surface for indirect prompt injection, where hidden instructions in the analyzed files could attempt to influence the agent's behavior. This is a known consideration for skills that process user-controlled documentation and is managed by platform-level guardrails.
Analysis of External Context: To fulfill its goal-clarity role, the skill instructs the agent to use tools like Read, Grep, or Glob to reconstruct project history. While this involves accessing file system data, the access is restricted to project-specific artifacts and is necessary for the skill's primary function of detecting goal drift.
No Executable Code Detected: The skill consists entirely of markdown instructions and metadata. It does not include or download any executable scripts, binaries, or third-party packages, which minimizes the risk of direct command injection or malicious code execution.

intent-keeper