miseon-stack/ai-skills-collection/map-agent-contracts
Claude Code and agent skill by miseon-stack/ai-skills-collection
- community
Checking that the SKILL.md file is still in the repository.
Identify every Agent actually visible in an AI product workflow and derive evidence-based I/O contracts, functional judgments, tool calls, global-context reads/writes, state, re-entry, and downstream handoffs. Use for multi-Agent product teardown from screenshots, browser pages, chats, canvas assets, task states, and errors, producing a standalone Agent-contract HTML report.
Install
npx vskill@latest install miseon-stack/ai-skills-collection/map-agent-contractsSKILL.md
The same SKILL.md was published earlier as miseon-stack/ai-skills-collection/map-agent-contracts.
Reverse Agent Contracts
Produce 02-agent-contracts.html unless the user requests another filename.
Non-negotiable boundaries
- Use only observable behavior. Public “thinking complete” or “planning complete” is a user-facing summary, not hidden chain-of-thought.
- A verbal plan is not a tool call. Confirm execution only with an asset, state transition, tool result, page result, or error.
- Do not assume a standard number of Agents. Record only names supported by page evidence.
- Distinguish a UI/session identity from a proven executing/orchestrating Agent.
- If no official tool name is visible, use a functional label and append
(功能命名,并非官方工具名). - Keep browser work read-only and follow the same privacy/safety restrictions as the journey skill.
- Tag claims as confirmed, inferred, suggested, or unknown.
Workflow
- Build an evidence ledger and chronological phase map.
- Identify each visible Agent name, first appearance, trigger, predecessor, successor, re-entry, and absent expected roles.
- For every Agent inspect six input sources: current user input, long-term user information, project global context, upstream Agent output, platform knowledge/public assets, and tool/runtime results.
- Summarize functional judgments only: completeness checks, confirmation gates, automatic continuation, stop conditions, modification, failure, validation, and completion.
- Inventory tools and classify each as confirmed result, announced-only, inferred function, or unknown.
- Inspect five output classes: user reply, page components, text/media assets, global-context writes, downstream task/handoff.
- Create one fixed-format ten-section contract card per Agent.
- Build global context, producer-consumer, version, reference, invalidation, and inconsistency tables.
- Draw the Agent input→judgment→tool→output→handoff flow. Use solid edges for explicit handoffs and dashed edges for inferred handoffs.
- List facts, inferences, suggestions, unknowns, and five highest-value validation questions.
Output contract
Read references/report-contract.md before writing. The file must be standalone UTF-8 HTML, responsive, searchable/scannable, and include evidence IDs in cards and diagrams.
Stop after the Agent-contract report. Do not write complete System Prompts in this skill.
Security
- Security scanPassed, 95 of 100Pattern checks on the skill's files.
- Model reviewPassed
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingskills/map-agent-contracts/SKILL.md
Popularity
- 7-day trend score
- 0
Not refreshed yet.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/miseon-stack/ai-skills-collection/map-agent-contracts)