The Agent Skills Directory

[SAFE]: The skill defines a prescriptive engineering workflow (Question, Research, Structure, Plan, Implement) intended to reduce hallucinations and ensure architectural integrity. It uses standard markdown files for documentation and templates.
[COMMAND_EXECUTION]: The skill includes two shell scripts (hooks/prompt-hook.sh and scripts/verify-session.sh).
hooks/prompt-hook.sh is designed to intercept prompts and inject status updates into the agent context based on the presence of QRSPI stage files in directories like .qrspi/ or .docs/. It performs localized file system checks and does not interact with the network or sensitive system paths.
scripts/verify-session.sh is a utility to ensure all five required stage files exist in a given directory.
[PROMPT_INJECTION]: While the skill uses strong directive language (e.g., "NON-NEGOTIABLE EXECUTION INVARIANTS"), this is used to enforce a specific engineering process rather than to bypass the agent's safety guidelines or override its core identity.

qrspi-methodology