docs/archive/2026-07-ai-artifacts/claude-skills/integration-pipeline-preview/02-classify-integration/SKILL.md
> **ARCHIVAL KNOWLEDGE — DO NOT USE OR REFERENCE.** > Superseded by `codemap/` + root `AGENTS.md`. Scheduled for deletion once the pipeline is verified (Phase 6). > Manifest: `docs/archive/2026-07-ai-artifacts/MANIFEST.md` --- name: 02-classify-integration description: Classifies the integration into a complexity tier via opus judgment, then persists it for the grounding loop. kind: process executor: haiku model: claude-haiku-4-5-20251001 --- # Classify integration Decides the integration's t
npx skillsauth add neohaskell/neohaskell docs/archive/2026-07-ai-artifacts/claude-skills/integration-pipeline-preview/02-classify-integrationInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
ARCHIVAL KNOWLEDGE — DO NOT USE OR REFERENCE. Superseded by
codemap/+ rootAGENTS.md. Scheduled for deletion once the pipeline is verified (Phase 6). Manifest:docs/archive/2026-07-ai-artifacts/MANIFEST.md
Decides the integration's tier (trivial / simple / moderate / complex / security-critical) using an opus-tier judgment leaf, then persists the decision to .integration-pipeline/classification.json. The grounding loop in phases 4, 5, 12, 13 reads this file to scale review intensity.
Classification is the load-bearing input for every grounding pass. A wrong call here cascades — a trivial-misclassified security integration skips the cryptographic and authorisation rubrics; a security-critical-misclassified mock test fixture accumulates a cascade of constEq / ScrubbedBytes / SHA-pinning recommendations. The decision step uses opus because the cost of a wrong call is higher than the cost of the model.
Outbound integrations almost always at least touch external IO (one of the cascade-rule signals), so the simple tier is rare; expect moderate for vanilla outbound HTTP / queue / file-storage integrations and security-critical whenever the integration handles secrets (API keys, OAuth tokens, webhook signing secrets).
./01-decide/SKILL.md and follow it on the available context (integration name, issue text, planned module path under integrations/Integration/<Name>, design draft if present, the diff scope so far). Verify: stdout contains a single JSON document with tier, rationale, and signals; tier is one of the five valid values../02-persist/SKILL.md and follow it with the JSON document from step 1. Verify: .integration-pipeline/classification.json exists with the chosen tier; pipeline.py status lists phase 2 as completed.Walk these steps in order. After each, run the verify check before continuing. If a verify fails, stop and surface.
security-critical on what the agent calls moderate) is recorded in the rationale field rather than silently substituted.development
> **ARCHIVAL KNOWLEDGE — DO NOT USE OR REFERENCE.** > Superseded by `codemap/` + root `AGENTS.md`. Scheduled for deletion once the pipeline is verified (Phase 6). > Manifest: `docs/archive/2026-07-ai-artifacts/MANIFEST.md` --- name: neohaskell-style-guide description: NeoHaskell coding style reference and enforcement rules. Load when writing, reviewing, or modifying any Haskell code in the NeoHaskell project. Triggers on 'NeoHaskell style', 'NeoHaskell conventions', 'how to write NeoHaskell', '
development
> **ARCHIVAL KNOWLEDGE — DO NOT USE OR REFERENCE.** > Superseded by `codemap/` + root `AGENTS.md`. Scheduled for deletion once the pipeline is verified (Phase 6). > Manifest: `docs/archive/2026-07-ai-artifacts/MANIFEST.md` --- name: neohaskell-security-review description: Security & Code Quality review for NeoHaskell. Use when reviewing code changes, PRs, or architectural decisions for security implications. Evaluates OWASP, NIST, EU compliance. Handles pipeline phases 2 (ADR review) and 10 (im
development
> **ARCHIVAL KNOWLEDGE — DO NOT USE OR REFERENCE.** > Superseded by `codemap/` + root `AGENTS.md`. Scheduled for deletion once the pipeline is verified (Phase 6). > Manifest: `docs/archive/2026-07-ai-artifacts/MANIFEST.md` --- name: neohaskell-qa-designer description: Testing QA Designer for NeoHaskell. Designs comprehensive test specifications with exhaustive edge cases, boundary conditions, and happy paths BEFORE implementation. Handles pipeline phase 6 (Test Spec Design). Outside-in TDD meth
development
> **ARCHIVAL KNOWLEDGE — DO NOT USE OR REFERENCE.** > Superseded by `codemap/` + root `AGENTS.md`. Scheduled for deletion once the pipeline is verified (Phase 6). > Manifest: `docs/archive/2026-07-ai-artifacts/MANIFEST.md` --- name: neohaskell-performance-review description: Performance review for NeoHaskell targeting 50k req/s throughput. Use when reviewing code for performance implications, INLINE pragmas, strictness, and allocation patterns. Handles pipeline phases 3 (ADR review) and 11 (imp