fixing-boot-failure

name	fixing-boot-failure
title	Fixing a Boot Failure
version	1
status	active
owners	[]
triggers	["boot failure","QEMU timeout","missing marker"]
prerequisites	["debugging-investigation","qemu-boot-testing","unsafe-rust-low-level-safety","agent-handoff"]
related_docs	["docs/development/debugging.md","tools/finnlib/qemu.py","docs/architecture/boot.md"]
category	Execution workflows
conditional_skills	["interrupt-exception-handling","uefi-bootloader-development","virtual-memory"]
implementation_gates	[]
related_milestones	["M0 Reproducible Build","M1 Dual-Architecture Boot"]
last_verified	{"base_commit":"d21a477","date":"2026-07-16","worktree_dirty":true,"context":"base commit plus uncommitted audit and agent-system worktree"}
description	Use when working on boot failure, QEMU timeout, missing marker; provides the FinnOS-specific fixing a boot failure workflow and evidence gates.

Fixing a Boot Failure

1. Name

fixing-boot-failure (Execution workflows; skill maturity: operational).

2. Purpose

Use a staged decision tree to localize image, firmware, loader, handoff, entry, paging, exception, emulator, or harness failures.

3. When to use this skill

Use for requests mentioning boot failure, QEMU timeout, missing marker, or when a dependency points to this skill. Load only after repository entry and before design or implementation.

4. When not to use this skill

Do not use this skill as a substitute for its adjacent subsystem skills or for evidence that the subsystem works. Do not load it only because a future FinnOS document mentions the subsystem.

5. Prerequisite skills

• debugging-investigation
• qemu-boot-testing
• unsafe-rust-low-level-safety
• agent-handoff

Read the full prerequisite closure in topological dependency order. If one cannot be satisfied, move the task to Blocked or Deferred; do not omit the dependency.

Conditional skills:

• uefi-bootloader-development for loader/handoff failures
• virtual-memory for paging failures
• interrupt-exception-handling for trap failures

Implementation gates are roadmap/runtime conditions, not additional documents to load automatically:

• No additional gate beyond the selected roadmap acceptance criteria.

6. Authoritative repository references

• docs/development/debugging.md
• tools/finnlib/qemu.py
• docs/architecture/boot.md

Re-read implementation and tests referenced by those documents. The documents establish intent/status boundaries, not runtime proof.

7. Current FinnOS context

Current x86 marker sequence is validated; ARM boot does not exist.

Registry verification used base commit d21a477 plus the dirty worktree context "base commit plus uncommitted audit and agent-system worktree" on 2026-07-16. This is not an integrated-revision claim. Reverify after HEAD, active PRs, or relevant source changes.

8. Required inputs

• User request or issue with desired outcome and architecture/profile scope.
• Current git status, recent history, active related issue/PR, and selected roadmap item.
• Relevant implementation, tests, invariants, ADRs, and exact baseline output.
• Toolchain/firmware/hardware versions when behavior crosses those boundaries.

9. Expected outputs

An evidence-backed fixing a boot failure result with scoped artifacts, tests, documentation, and handoff. Include acceptance evidence, residual limitations, and next dependency rather than only code.

10. Step-by-step workflow

1. Verify image files/manifest
2. Verify firmware/QEMU command
3. Classify last loader marker
4. Validate BootInfo and kernel entry
5. Inspect CR3/mappings/exceptions
6. Check harness status/ordering/timeout
7. Add regression at owning layer
8. Run the narrow regression, then all required subsystem/repository checks.
9. Update canonical docs/status and finish the agent-handoff template.

11. Repository-specific commands

./tools/finn check

Run commands from the repository root. A command listed here is a baseline/gate, not evidence that absent future functionality has a runnable target.

12. Architecture considerations

State shared semantics explicitly; isolate x86-64 and ARM64 mechanisms.

State guest architecture separately from host architecture and emulator model. Maintain a parity row for changed semantics and document intentional differences.

13. Safety constraints

Preserve the boundary described by the current state: Current x86 marker sequence is validated; ARM boot does not exist. Apply .agents/checklists/pre-change.md; for kernel, driver, security, architecture, or UI work also apply the matching checklist.

14. Testing requirements

Make Add regression at owning layer observable with a negative case, then run the narrow and aggregate gates. Do not permanently hard-code test counts; counts belong to dated evidence reports.

15. Documentation requirements

Update the canonical behavior/status document, relevant architecture/reference material, test/build instructions, limitations, and this skill registry if any command, gate, or current-context statement changes.

16. Review checklist

• The workflow proves Verify image files/manifest before relying on downstream assumptions.
• The observed baseline and first failing/divergent point are recorded.
• Every prerequisite and roadmap dependency is satisfied or explicitly marked blocked.
• Bounds, ownership, rollback, error paths, and resource limits were reviewed.
• x86-64 and ARM64 semantics are mapped without unsupported parity claims.
• Negative/failure tests exist at the layer that owns the behavior.
• Commands, logs, artifact/profile/architecture, and results are preserved.
• Canonical documentation and affected agent skills agree with behavior.

17. Completion criteria

• All stated outputs exist and remain within the task's scope/non-goals.
• The narrow regression and required repository/subsystem gates pass.
• Acceptance criteria are tied to source, test, runtime, or hardware evidence as appropriate.
• Unknowns and unsupported environments remain explicit; no planned feature is promoted.
• The final diff is reviewed and a structured handoff is complete.

18. Common failure modes

• Starting from an audit statement instead of re-reading changed source and tests.
• Using a successful compile or marker as evidence for a broader subsystem claim.
• Ignoring this skill's current constraint: Current x86 marker sequence is validated; ARM boot does not exist.
• Changing shared policy while testing only one architecture or one happy path.
• Losing the exact failing artifact/log by rebuilding before capture.

19. Forbidden shortcuts

• Do not weaken, delete, reorder, or broaden validators merely to make a test pass.
• Do not claim ARM64, physical hardware, userspace, Peony, security, or release support without its own acceptance evidence.
• Do not bypass a named prerequisite by embedding a temporary incompatible abstraction.
• Do not commit target/, build/out/, firmware, credentials, local paths, or unreviewed generated output.
• Do not disregard the roadmap/status gate: Do not use this skill as a substitute for its adjacent subsystem skills or for evidence that the subsystem works.

20. Handoff requirements

Use .agents/templates/handoff-template.md. Include objective, starting/final Git state, task state, skills used, files, exact commands/results, evidence classification, docs/status changes, unknowns, blockers, risks, next action, and next skills. Distinguish locally verified worktree changes from integrated behavior.

21. Examples

Request: "Work on boot failure." Correct response: begin by verify image files/manifest, then verify firmware/qemu command, and require evidence for add regression at owning layer before changing status. Incorrect response: create a plausible subsystem scaffold and mark the roadmap item complete because it compiles.

22. Skill maintenance notes

Canonical source: .agents/scripts/skill_registry.py. Increment version for material policy/workflow/gate changes, update last_verified only after reinspection, run python3 .agents/scripts/render_skills.py, review generated diffs, then run python3 .agents/scripts/validate.py --all. Follow .agents/GOVERNANCE.md; never hand-edit generated skills.