No-Input Follow-up Review

Review

Correct

• All requested follow-up artifacts exist. I inspected:
  • funding-drilldown-L2-results/baseline-hours-investigation.md
  • funding-drilldown-L2-results/evidence-hardening.md
  • funding-drilldown-L2-results/code-forensics-risky-aps.md
  • funding-drilldown-L2-results/patch-readiness-matrix.md
  • funding-drilldown-L2-results/next-todos-summary.md
  • funding-drilldown-L2-results/no-input-followup-summary.md
• No main funding docs were edited. git status --short docs/1-projects/bsfz-funding shows only untracked follow-up artifacts under funding-drilldown-L2-results/; no diffs exist for STRATEGY_public-funding.md, funding-work-packages.md, bsfz-antrag-vorhaben.md, bsfz-ablehnungsrisiken.md, or funding-application-plan.md.
• Baseline-hour recommendation is justified and conservative. baseline-hours-investigation.md:24-38 reconstructs the current AP-catalog maximum as 1.760 + 100 + 50 + 60 + 1.220 + 640 + 340 + 100 = 4.270h; I verified the cited source table in funding-work-packages.md:286-294. The stale 4.110h references are correctly traced to L1/L2-A and strategy submission ranges (baseline-hours-investigation.md:13-18, verified against funding-drilldown-L1.md:10, :95, funding-drilldown-L2-A.md:53, and STRATEGY_public-funding.md:134-136). The recommendation explicitly limits 4.270h to a maximum/upper bound, not a year-1 submission recommendation (baseline-hours-investigation.md:64-75, :99-112).
• Line/source references are accurate enough for the reviewed purpose. Spot checks confirmed the key references:
  • AP1 source/risk lines in evidence-hardening.md:25 and code-forensics-risky-aps.md:30-35 match funding-work-packages.md:19-24, :32-46, apps/api/docs/tax-engine.md:83-140, and adr-012-toms-tax-deferral.md:32-41, :58-64.
  • AP18/AP19/AP24/AP26 conditions in evidence-hardening.md:29-33 and code-forensics-risky-aps.md:65-179 match the documented AP constraints in funding-work-packages.md:189-204, :240-265 and bsfz-ablehnungsrisiken.md:236, :249-251, :283-285.
  • The workflow/n8n caution for AP26 is supported by workflow-orchestration.md:21-40, :67-87, including explicit n8n prototyping and migration language.
• Evidence statuses are conservative. The legend distinguishes belegt, inferiert, fehlend, Berater-Validierung, and Code/Git-Prüfung nötig (evidence-hardening.md:3-9). High-impact claims are not over-promoted: V2 breadth and hybrid submission are marked inferred or mixed (evidence-hardening.md:17-19); F&E splits and audit/widerspruch topics remain beraterpflichtig (evidence-hardening.md:26, :42-43).
• Code/git forensics do not overclaim. The risky AP forensics repeatedly distinguish document evidence from implementation evidence:
  • AP1 is kept but with weak/retrospective chronology warning (code-forensics-risky-aps.md:51-61).
  • AP18 is blocked without CRDT technology/ADR/spike (code-forensics-risky-aps.md:73-89).
  • AP19 is blocked without a custom offline validation protocol (code-forensics-risky-aps.md:105-119).
  • AP24 is conditioned on an Edge-ML/On-Device-ML prototype or ADR (code-forensics-risky-aps.md:135-151).
  • AP26 is only bedingt, with explicit n8n/standard-automation risk (code-forensics-risky-aps.md:165-179). I also ran a repository search excluding funding-result artifacts for yjs, automerge, crdt, tensorflow, tfjs, and onnx; it returned no implementation/dependency hits, supporting the cautious AP18/AP24 findings.
• Patch-readiness categories are actionable and safe. patch-readiness-matrix.md:17-56 cleanly separates safe now, needs FZulG/Steuerberater validation, needs code evidence, defer, and discard. The batch plan keeps only consistency/label patches in Batch 1 and reserves Antragstexte/F&E splits for code evidence + berater/user validation (patch-readiness-matrix.md:58-89).
• Next TODOs separate agent-work from user/berater decisions. next-todos-summary.md:22-31 says AP1 research can continue without user input but final Antragssprache needs user decision; :35-61 correctly assigns AP18/AP19/AP24/AP26 technology choices and berater involvement to user/berater decisions; :94-106 explicitly lists the remaining user decisions and recommends asking for Batch 1 approval first. no-input-followup-summary.md:79-85 repeats the decision list clearly.

Blocker

• None for the review target. All requested follow-up artifacts exist, and I found no unauthorized main-document edits.

Note

• The delegated read inputs docs/1-projects/bsfz-funding/plan.md and docs/1-projects/bsfz-funding/progress.md were absent during this review. The required review artifacts themselves were present, so this did not block the artifact review.
• Minor bookkeeping inconsistency: baseline-hours-investigation.md:7, no-input-followup-summary.md:11, and root progress.md mention missing context.md/plan.md; this appears inherited from the producing task rather than the current review request, which named plan.md/progress.md. It does not affect the funding conclusions, but future summaries should use the exact requested filenames to avoid confusion.
• safe now in patch-readiness-matrix.md:19-23 still correctly depends on User-Freigabe; it should not be interpreted as permission to edit main funding documents automatically.