# Agent Failure Archive

> 186 real AI agent post-mortems, 107 of them measurement failures. Free tools, paid via x402.

- name: io.github.HanbeenMoon/agent-failure-archive
- version: 0.3.1
- connection class: R0 (autonomous)
- trust: 94/100
- quality: 79/100 (good)
- owner: not claimed
- descriptor: https://api.protogrid.dev/v1/servers/io.github.HanbeenMoon%2Fagent-failure-archive
- tools: https://api.protogrid.dev/v1/servers/io.github.HanbeenMoon%2Fagent-failure-archive/tools

**An agent can connect right now, no human step.** Remote endpoint, no authentication, reachable on the last probe.

## Connection (mcpServers)

```json
{
  "mcpServers": {
    "agent-failure-archive": {
      "type": "http",
      "url": "https://desktop-ai2ata5-1.tailfeb765.ts.net/mcp"
    }
  }
}
```


## Trust

- hygiene: 100
- liveness: 99
- freshness: 85
- provenance: 90
- drivers: +repository +namespace-matches-repo +website +reachable +uptime-98% ~updated-56d-ago
- Derived from observable signals (official registry feed and our own credential-free probes); no code audit performed.

## Quality 79/100 (good)

### protocol: 75 (weight 25)

- warn `protocol.modern`: newest supported version is 2025-11-25; 2026-07-28 not supported, older versions are deprecated until 2027-07-28
- n/a `protocol.stateless`: only applies to 2026-07-28 servers
- pass `protocol.transport`: streamable HTTP
- n/a `protocol.list_ttl`: only applies to 2026-07-28 servers

### authorization: n/a (weight 15)

- n/a `auth.prm`, `auth.as_metadata`, `auth.cimd`: no remote uses OAuth
- n/a `auth.secret_in_url`: no templated URL

### tool hygiene: 71 (weight 30)

- pass `tools.descriptions`: every tool has a description
- pass `tools.description_length`: descriptions are concise
- pass `tools.schemas`: every tool has an object input schema
- fail `tools.annotations`: no tool declares readOnlyHint or destructiveHint, so clients cannot tell safe calls from risky ones
- fail `tools.directory_hints`: no tool declares all four of readOnlyHint, destructiveHint, idempotentHint and openWorldHint; missing: brief (readOnlyHint, destructiveHint, idempotentHint, openWorldHint); archive (readOnlyHint, destructiveHint, idempotentHint, openWorldHint); search (readOnlyHint, destructiveHint, idempotentHint, openWorldHint); …
- pass `tools.token_cost`: about 578 tokens to load every tool
- pass `tools.api_dump`: tools are not a one-to-one API dump

### stability: 100 (weight 15)

- pass `stability.changes`: no tool changes in 30 days
- pass `stability.rug_pull`: no tool changed its meaning under the same name

### dependencies: n/a (weight 15)

- n/a `deps.known_vulns`, `deps.mcp_sdk_version`, `deps.resolvable`: no npm or PyPI package

- tools: 9, about 578 tokens to load them all
- tool set hash: d8521b3b142d820dd6a75b0f8553a53b
- badge: [![protogrid quality](https://protogrid.dev/badge/io.github.HanbeenMoon/agent-failure-archive.svg)](https://protogrid.dev/servers/io.github.HanbeenMoon/agent-failure-archive)
- Checks run on what our credential-free, read-only probes observe; tools are never called and no code audit is performed.

### Tool changes

_No tool definition changes recorded._

## Remotes

- https://desktop-ai2ata5-1.tailfeb765.ts.net/mcp (streamable-http, auth none, reachable true, uptime30d 0.98275864)

## Packages

_none_

## Tools (9)

- `archive`: $1.00. Every case in one response. One payment, no subscription, no account.
- `audit`: $0.02. Full audit of a claim: why each tripped check matters, the incident behind it with measured numbers, and what to run.
- `brief`: $0.05. Pre-flight risk brief before an irreversible action, drawn from the ways that class of action actually failed.
- `catalog`: Free, no wallet. What the archive holds, what each paid tool costs, and how payment works.
- `contents`: Free, no wallet. Every case title in the archive, tagged with the trap it illustrates. Titles only, no bodies. Read this to see what the paid archive actually contains before deciding it is worth a dollar.
- `precheck`: Free, no wallet. Check a conclusion against nine known ways of fooling yourself and get back which checks it trips plus the question each one asks. Use it before writing 'we found that'. The paid audit tool adds why each matters, the real incident with the numbers measured at the time, and what to run.
- `research`: $0.25. The 107 measurement failures from an eight-month attempt to measure one person's individuation with embeddings.
- `sample`: Free, no wallet. Two complete post-mortems from the archive.
- `search`: $0.01. Three real agent post-mortems matching a symptom, each with root cause, the fix that worked, and the prevention rule.