neohaskell/docs/archive/2026-07-ai-artifacts/claude-skills/feature-pipeline-preview/11-build-loop/03-fix-iter

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 Superseded by codemap/ + root AGENTS.md.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: 03-fix-iter description: Reads build and test logs, identifies the root cause, and edits implementation files to fix it. kind: leaf executor: sonnet model: claude-sonnet-4-6

Fix Iter

Reads the latest build and test logs and applies a targeted fix to the implementation. Test files are immutable.

Inputs

• .pipeline/build.log
• .pipeline/test.log
• docs/architecture/<adr-number>-<slug>.md — architecture doc.

Plan

1. Read both logs → verify: at least one ends with a failure marker.
2. Identify the root cause (compile error, type mismatch, test failure, runtime exception) → verify: a single hypothesis is named.
3. Edit only implementation files → verify: no file under core/test/ or testbed/tests/ is touched.
4. Re-run nix develop --command cabal build all (or full loop on caller's next step) → verify: at least one previously-failing signal now passes.

Assumptions:

• Test files are immutable; only implementation files may change.
• The iteration counter caps at 10. After the 10th iteration with failure, refuse and surface for the maintainer.
• NeoHaskell style is strict — fixes must respect pipes, do+let, case...of, qualified imports.
• Boy scout rule applies. When the fix re-opens a file, any pre-existing style debt in that file is cleaned up in the same diff — unqualified imports, point-free top-levels, wildcard _ params, let..in/where, $, <>/++, raw String/IO/Either. Scoped strictly to files the fix is already editing; nothing else gets reformatted. See ../../references/nhcore-context.md#boy-scout-rule.

If any assumption fails, refuse — do not guess.

Steps

1. Load the build and test logs and the architecture doc.
2. Locate the first failure in the logs and identify the responsible source file.
3. Form a single hypothesis for the root cause.
4. Edit the implementation file(s) to address the hypothesis.
5. Re-run nix develop --command cabal build all > .pipeline/build.log 2>&1.
6. If build passes, the caller will re-run tests on the next loop turn.
7. Increment the iteration counter via python3 .claude/skills/feature-pipeline-preview/scripts/pipeline.py iter 11.
8. If the counter exceeds 10, refuse and escalate.

Output

Edited implementation file(s); rebuilt .pipeline/build.log; iteration counter advanced.

Refusals

• Iteration counter > 10 → refuse: "build loop exhausted; escalate to maintainer".
• Any test file modified → refuse: "tests are immutable".
• No identifiable failure in either log → refuse: "no failure found; loop should not be running".