Prompt file imported from somster/agentic-controls-harness (
.github/prompts/generate-controls.prompt.md). Fill in{{playbook}}before use. Copyright stays with the author.
Create implementation-ready controls for {{playbook}}.
Review the Mitigation Strategies section in docs/Agentic-AI-Threats-and-Mitigations.pdf first, then produce output in the structure and tone of docs/playbook-template-security-controls.md.
Before extracting from the PDF with Python tools, ensure the workspace Python environment is configured and the required PDF package is available in that environment; otherwise use an available read-only CLI extractor.
Use {{playbook}} as the target playbook identifier, title, or output filename. State assumptions only when needed.
Focus on the mitigations relevant to {{playbook}} and turn them into concrete implementation controls with code or config examples where useful.
Always write the playbook to a file under docs/.
If {{playbook}} already looks like a path under docs/, write the playbook there.
Otherwise, generate a safe slug filename under docs/ and write the output there (for example docs/playbook2-memory-poisoning-ai-knowledge-corruption.md).
After writing, return the final file path and a brief summary in chat.
Example invocations:
/generate-controls playbook1-langchain-controls.md/generate-controls playbook 2 - tool execution confinement/generate-controls playbook3-vector-store-controls.md