Skip to content

AGENTS.md

Latest commit

 

History

History
112 lines (83 loc) · 6.67 KB

AGENTS.md

File metadata and controls

112 lines (83 loc) · 6.67 KB

<<<USABLE_MCP_SYSTEM_PROMPT_START>>>

🧠 Usable MCP - SYSTEM PROMPT (LONG-TERM MEMORY)

This is your main way of storing and fetching data. Always consult it before starting a task and whenever you need more context.

📚 Repository Documentation Maintenance

CRITICAL: The following authoritative knowledge fragments document this repository and MUST be kept up to date:

  1. "Usable PR Validator - GitHub Action Documentation" (Fragment ID: a859c565-ddb9-4d3e-b716-4b644b08e161)

    • Purpose: Complete documentation of what this repo is, its architecture, features, and how to use it
    • When to Update: Any time features, inputs, outputs, or usage patterns change
    • Content: Overview, features, configuration, usage examples, troubleshooting

  2. "Usable PR Validator - Deployment & Release Process" (Fragment ID: d2ca8a68-fb9e-4098-a4ac-5eb04013a9fd)

    • Purpose: Documents deployment process, release automation, and CI/CD pipeline
    • When to Update: Any time release process, CI/CD workflows, or versioning strategy changes
    • Content: Release workflow, GitHub Marketplace publishing, testing pipeline

Your Responsibilities

When making changes to this repository, you MUST:

  1. Search for these fragments at the start of any significant work session

  2. Review their content to ensure accuracy before making changes

  3. Update fragments immediately when you:

    • Add/remove/modify action inputs or outputs
    • Change the release process or CI/CD pipeline
    • Update core functionality or architecture
    • Fix critical bugs or security issues
    • Add new features or examples
    • Modify deployment procedures

  4. Use update-memory-fragment with the correct fragment ID

  5. Preserve fragment structure while updating outdated information

  6. Add timestamps or version references to track when updates were made

Linking Strategy

These two fragments are conceptually linked and should reference each other:

  • The main documentation fragment should reference deployment for release/version info
  • The deployment fragment should reference main docs for feature context
  • Both should be tagged with repo:usable-pr-validator for easy discovery

Search Query Examples:

"Usable PR Validator documentation"
"deployment process for usable-pr-validator"
"repo:usable-pr-validator github action"

Detailed instructions for each tool are embedded in its MCP description; read them before you call the tool.

Search Discipline

  • Start or resume every task with agentic-search-fragments and rerun whenever scope expands or you lack certainty.
  • Provide workspace scope and begin with repo:usable-pr-validator tags; iterate until the tool reports decision: "SUFFICIENT".
  • If the agentic tool is unavailable, fall back to search-memory-fragments, then return to the agentic loop as soon as possible.
  • Respect the tool's safety rails—if you see invocationLimitReached: true, stop rerunning the tool and document the uncovered gap instead. Reset the attempt counter whenever you start a materially different search objective.
  • Use get-memory-fragment-content for deep dives on selected fragment IDs and cite titles plus timestamps in your responses.

Planning Loop

  • Plan: Outline sub-goals and the tools you will invoke.
  • Act: Execute tools exactly as their descriptions prescribe, keeping actions minimal and verifiable.
  • Reflect: After each tool batch, summarise coverage, note freshness, and decide whether to iterate or escalate.

Verification & Documentation

  • Verify code (lint, tests, manual checks) or obtain user confirmation before relying on conclusions.
  • Capture verified insights by using create-memory-fragment or update-memory-fragment; include repository tags and residual risks so the team benefits immediately.

Freshness & Escalation

  • Prefer fragments updated within the last 90 days; flag stale sources.
  • If internal knowledge conflicts or is insufficient after 2–3 iterations, escalate to external research and reconcile findings with workspace standards.

Repository: usable-pr-validator WorkspaceId: 60c10ca2-4115-4c1a-b6d7-04ac39fd3938 Workspace: Flowcore Workspace Fragment Types: knowledge, recipe, solution, template, feature request, instruction set, llm persona, llm rules, plan, prd, research

Fragment Type Mapping

The following fragment types are available in this workspace:

  • Knowledge: 04a5fb62-1ba5-436c-acf7-f65f3a5ba6f6 - General information, documentation, and reference material
  • Recipe: 502a2fcf-ca6f-4b8a-b719-cd50469d3be6 - Step-by-step guides, tutorials, and procedures
  • Solution: b06897e0-c39e-486b-8a9b-aab0ea260694 - Solutions to specific problems and troubleshooting guides
  • Template: da2cd7c6-68f6-4071-8e2e-d2a0a2773fa9 - Reusable code patterns, project templates, and boilerplates
  • Feature Request: d016c715-0499-4af5-b69b-950faa4aa200 - A Feature request for products we develop, these should be tagged by the repo it is tied to and the product name
  • Instruction Set: 1d2d317d-f48f-4df9-a05b-b5d9a48090d7 - A set of instructions for the LLM to perform a set of actions, like setting up a project, installing a persona etc.
  • LLM Persona: 393219bd-440f-49a4-885c-ee5050af75b5 - This is a Persona that the LLM can impersonate. This should help the LLM to tackle more complex and specific problems
  • LLM Rules: 200cbb12-47ec-4a02-afc5-0b270148587b - LLM rules that can be converted into for example cursor or other ide or llm powered rules engine
  • Plan: e5c9f57c-f68a-4702-bea8-d5cb02a02cb8 - A plan, usually tied to a repository
  • PRD: fdd14de8-3943-4228-af59-c6ecc7237a2c - A Product requirements document for a project or feature, usually targeted for a repository
  • Research: ca7aa44b-04a5-44dd-b2bf-cfedc1dbba2f - Research information done with the express purpose of being implemented at a later date.

Fragment Type Cheat Sheet

  • Knowledge: reference material, background, concepts.
  • Recipe: human step-by-step guides and tutorials.
  • Solution: fixes, troubleshooting steps, postmortems.
  • Template: reusable code/config patterns.
  • Instruction Set: automation workflows for the LLM to execute.
  • Plan: roadmaps, milestones, "what/when" documents.
  • PRD: product/feature requirements and specs.

Before choosing, review the workspace fragment type mapping to spot custom types that may fit better than the defaults.

Quick picker: “How to…” → Recipe · “Fix…” → Solution · “Plan for…” → Plan · “Requirements…” → PRD · “What is…” → Knowledge · “Reusable pattern…” → Template · “LLM should execute…” → Instruction Set.

<<<USABLE_MCP_SYSTEM_PROMPT_END>>>