---
name: H-Diagnose
slug: h-diagnose-2
category: Quality
description: H-Diagnose stabilizes a concrete failure, generates distinct rival explanations, tests discriminating observations read-only in parallel, and ranks them by evidence while retaining losing hypotheses for later use.
github: "https://github.com/m0n0x41d/haft/tree/main/packages/haft-pi/skills/h-diagnose"
language: Go
stars: 1383
forks: 102
install: "npx degit https://github.com/m0n0x41d/haft/tree/main/packages/haft-pi/skills/h-diagnose ~/.claude/skills/h-diagnose"
installs_to: ~/.claude/skills/h-diagnose
source_path: packages/haft-pi/skills/h-diagnose/SKILL.md
collection_size: 24
category_size: 1354
collection_url: "https://dirskills.com/collections/m0n0x41d/haft"
added: 2026-08-19T07:27:49.042Z
last_synced: 2026-08-19T07:27:49.042Z
canonical_url: "https://dirskills.com/skills/h-diagnose-2"
---

# H-Diagnose

H-Diagnose stabilizes a concrete failure, generates distinct rival explanations, tests discriminating observations read-only in parallel, and ranks them by evidence while retaining losing hypotheses for later use.

**Install:**

```bash
npx degit https://github.com/m0n0x41d/haft/tree/main/packages/haft-pi/skills/h-diagnose ~/.claude/skills/h-diagnose
```

## README

# h-diagnose — Test rival explanations

Retrieve current FPF source only when pattern choice is material to the
diagnosis. Mechanical reproduction, logs, and direct implementation probes do
not need a ritual query. Inspect a known SourceID/UnitID directly; otherwise
use `mode="concern"` and treat the returned `candidate_set` as incomplete
navigation, not a selected pattern. Before relying on one candidate, inspect
its exact identifier and direct pattern body. Keep several candidates live or
abstain when the returned basis is insufficient. Never run a query after the
diagnosis merely to manufacture source support for an already-chosen story.

Keep symptom, hypothesis, probe, observation, and verdict distinct. Include a
rival that challenges the initial framing. Prefer safe parallel probes; label
design-time inference separately from runtime evidence. Keep losing hypotheses
with return conditions. Persist only on explicit request or when a named
receiving use needs replay.