Recover and attest target-linked TH08 VC7 CRT, compiler-runtime, and D3DX library functions after authored coverage. Use for library inventory boundaries, archive/member provenance, hash pinning, relocation-aware comparison design, or whole-link preparation; do not use for repository third-party submodules.
Compile TH08 functions with the repository VC7 build and compare configured COFF functions against the hash-attested 1.00d target using config/match-units.toml and scripts/compare-function.py. Use for focused code-generation tuning, relocation diagnosis, or exact-match verification.
Reconstruct bounded functions from the original Japanese TH08 1.00d executable using hash-attested target evidence, upstream GensokyoClub source, and explicitly labeled TH06/TH07 corroboration. Use for TH08 disassembly, ABI recovery, function naming, layout recovery, source migration, or implementation work.
Replace raw TH08 object offsets, anonymous fields, and absolute field views with evidence-backed C++ types and names while preserving accepted VC7 bytes and playable modern-port behavior. Use for semantic cleanup of already-authored source; do not use for new function recovery or target-linked library work.
Generate and interpret target-pinned TH08 instruction and ABI fact packets with scripts/typed-re.py. Use for stack layout, register homes, access widths, direct calls, return cleanup, or VC7 source-shape diagnosis before strict comparison.
Instructions for N0zoM1z0/th08, covering th08 reconstruction agent rules, before changing reconstruction state, evidence and status rules, analysis database safety and abi and implementation.
AGENTS.md instructions for N0zoM1z0/touhou-university, covering touhou university — codex entry point, read before editing, non-negotiable product rules, non-negotiable engineering rules and working loop.
Turn a disputed, current, or consequential claim into a source-backed brief by decomposing it, finding the closest authoritative evidence, checking dates and independence, and calibrating the conclusion. Use for fact-checking, rumor verification, news claims, provenance questions, or briefs that need links and…
Strip an overcomplicated plan to a working design with at most three moving parts, expose the first real constraint that breaks it, and add complexity back only when justified. Use for architecture sprawl, tool overload, premature scaling, tangled workflows, or plans nobody can explain simply. Do not use to dismiss…
Review conduct, content, code, or decisions against an explicit rule set by establishing authority and scope, separating facts from allegations, testing each required element, grading severity, and recommending a proportionate corrective action. Use for policy compliance, specification conformance, governance…
Append truthful Assisted-by trailers when preparing, creating, or explicitly amending a Git commit whose staged changes were materially shaped by one or more Gensokyo character skills. Use for commit-message construction after a character workflow changed the commit's artifact, evidence, or decision. Do not use for…
Discover plausible inputs, state transitions, orderings, and dependency behaviors that violate an expected contract, prove the smallest genuine surprise, and forge it into a deterministic regression test. Use for edge-case design, negative testing, failure injection, fragile happy paths, or bugs that need a durable…
Convert an uncertain idea into the smallest runnable experiment, reuse relevant prior art, observe a decisive signal, and scale only after evidence. Use for spikes, proof-of-concepts, feasibility questions, unfamiliar libraries, or competing implementation approaches. Do not use when the main need is incident…
Reconstruct how an existing artifact or system works by inventorying observable surfaces, mapping interfaces and flows, forming competing mechanism hypotheses, probing them safely, and building a minimal explanatory replica. Use for undocumented formats, protocols, binaries, devices, legacy behavior, compatibility…
Isolate the real abnormal condition in a noisy incident, reduce it to the smallest relevant scope, and choose a minimal reversible recovery with explicit verification. Use for regressions, outages, flaky failures, broken workflows, or messy bug reports with many symptoms and competing explanations. Do not use for…
Convert a risky multi-step operation into a dependency-ordered runbook with observable checkpoints, irreversible-action gates, honest rollback or compensation paths, abort conditions, and end-to-end verification. Use for deployments, migrations, cutovers, key rotations, bulk transformations, or maintenance where…
Expose hidden assumptions in a consensus, strategy, requirement, or design by stating the default precisely, inverting one assumption at a time, constructing the strongest falsifiable counterexample, and reconciling the result with evidence. Use for adversarial reframing, pre-mortems, stuck debates, suspicious…
Decompose one bounded deliverable into genuinely independent work packets, give each packet an ownership and evidence contract, and recombine the results through an explicit merge order and integrated acceptance check. Use for parallel contributors, subagents, multi-package changes, or large work with real dependency…
Diagnose systems by mapping interfaces, ownership, trust zones, lifecycle transitions, and edge conditions, then trace failures across the most suspicious boundary. Use for integration bugs, distributed systems, API mismatches, unclear responsibility, state transitions, migrations, or failures that appear only between…