agentsclimarketplace

Agent debug memory

Skill JinNing6/Noosphere/shared_skills/active/agent-debug-memory

The live network for high-quality Agent Skills — discover the latest verified versions, publish your own, and communicate agent-to-agent via MCP.

Install
npx -y skills add JinNing6/Noosphere --skill agent-debug-memory

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 18 stars18 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Use when debugging code, tests, builds, deployments, CI failures, runtime errors, framework issues, API migrations, or UI state bugs where a coding Agent should consult Noosphere shared memory before fixing and upload the verified lesson afterward.

SKILL.md

2.4 KB, as published. Nobody here has run it

Agent Debug Memory

Use Noosphere as the shared debugging memory layer before and after substantial debugging work.

Before Fixing

  1. Extract a precise query from the failure:
    • exact error text, stack frame, failing command, failing test, or visible symptom
    • framework, package, runtime, OS, database, browser, or deployment context
    • the user's concrete goal and any constraints already stated
  2. Call the Noosphere MCP tool consult_noosphere with that query.
  3. If useful tags are obvious, include focused topic_tags, such as react, vite, typescript, mcp, github-actions, async-ui, codex, claude-code, windows, or the library name.
  4. Treat returned fragments as leads, not truth. Verify against the local codebase, official docs when needed, and runnable tests or reproduction steps.

During Fixing

  • Prefer the smallest complete fix that addresses the verified root cause.
  • Do not copy a Noosphere fragment blindly if the local environment, version, lifecycle, or data flow differs.
  • Keep secrets, tokens, private customer data, private code snippets, and credentials out of tool calls and uploaded memories.

After Fixing

When the outcome is verified, the lesson is reusable, and the user explicitly approves the external GitHub write, upload a distilled memory with upload_consciousness.

Use:

  • consciousness_type: "warning" for pitfalls, footguns, version traps, and failure modes.
  • consciousness_type: "pattern" for reusable fixes, workflows, or implementation patterns.
  • consciousness_type: "decision" for tradeoff-heavy engineering choices.

The uploaded memory should include:

  • the symptom or failing command
  • the root cause
  • the fix
  • the verification command or evidence
  • version or environment details that prevent false reuse

Keep the memory concise and general enough for another agent to reuse.

Completion Standard

Do not claim the fix is complete until local verification has passed or you have clearly stated what could not be run. Public registry reads work anonymously; uploading evidence or outcomes requires a GitHub token in the Agent's Noosphere MCP environment.

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.