idea-spark

name	idea-spark
description	Generate ONE reviewer-defensible, implementable research idea with a concrete method and falsification plan from a stated research direction. Use when the user asks for a research idea, novelty analysis, bottleneck diagnosis, or paper-shape suggestion. Skip code review, debugging, and unconstrained brainstorming without research context.

Idea Spark Skill

Convert an under-specified research direction into ONE reviewer-defensible Oral-level research proposal — grounded in 1947 ICLR/ICML/NeurIPS papers (2021-2025) — via a 5-phase workflow: retrieve recent literature, diagnose the bottleneck, select + generate a candidate using corpus-derived ideation pattern cards, run it through a quality gauntlet, expand into an idea card.

This file is the operational runbook. Design rationale (the 7 design principles, why each contract is shaped this way, removed-check history) lives in references/design-notes.md — read it only when modifying or evaluating the skill, never needed to run it. First-time installation lives in references/setup.md.

When to use

• "Give me a research idea in {area} I could pursue." / "What's the most impactful next step in this direction?"
• "Help me sharpen this vague direction into an Oral-level proposal."
• "What's the bottleneck of this problem?" / "Run a novelty audit on this idea."

When NOT to use

• Code review, debugging, refactoring. Summarizing one paper. Cross-decade survey writing.
• Free-association brainstorming with no research context. Engineering integration tasks ("ship this feature in our system").
• Pure benchmark / dataset construction work — the 15-pattern vocabulary handles benchmark audit (controlled_diagnostic_design) but not benchmark construction.

Setup (first use only)

Follow references/setup.md. Quick version — set two shell variables, install deps, verify:

SKILL_DIR=<path to this folder>                  # e.g. ~/.claude/skills/idea-spark (Claude Code), ~/.codex/skills/idea-spark (Codex CLI), or any clone location
RUN_DIR="$PWD/ideaspark_run/<topic-slug>" && mkdir -p "$RUN_DIR"   # convention below; any absolute dir works
python3 -m pip install feedparser openreview-py beautifulsoup4 pymupdf
python3 "$SKILL_DIR/scripts/run.py" check_connectors   # from the SAME shell you'll run phases from

Credentials go in .env (OpenReview user/pass + Semantic Scholar key — see setup.md); the orchestrator auto-loads it. Optional: xelatex/tectonic for PDF cards.

Research-pipeline handoff

收尾时，把最终 idea 卡片复制/汇总到 research_run/<课题slug>/stage1_ideas/ 下，并写入或追加 research_run/<课题slug>/stage1_ideas/idea_cards.md 索引。脱离 research-pipeline 单独调用时，退当前工作目录。

---

How to run: the next loop

The canonical way to drive a run is the run-state navigator:

python3 "$SKILL_DIR/scripts/run.py" next --dir "$RUN_DIR" --query "<user's research question>"

Run-dir convention (one run = one directory, named by the host BEFORE the first command): $PWD/ideaspark_run/<topic-slug> — a short kebab-case slug distilled from the user's direction (e.g. ideaspark_run/diffusion-watermark); if the slug is taken, append _2, not a timestamp. NEVER reuse a directory that already contains a phase0/ — every phase writes into $RUN_DIR and would clobber the prior run. The skill itself never names the dir (any absolute path works); this convention exists so runs from different harnesses land in one predictable place instead of each agent improvising.

next inspects the artifacts already on disk and prints EXACTLY one next step — either a Bash command to run verbatim, or an LLM sub-agent spec (system-prompt path + input file paths + output path + the routing signal to report back). It is read-only and idempotent (safe to re-run anytime, including to resume an interrupted run). The host loop is:

1. Run next.
2. Do what it says (bash → run the command; llm_subagent → execute in an ISOLATED context per the Context discipline rules below).
3. Run next again. Repeat until it reports a terminal state (DONE, do_not_generate, or phase_3_failed).

next encodes the full phase graph — the mandatory full-text gate, the citation gate, the abandon-retry branch, the falsification re-audit branch, and the correct Phase 4 flags per path — so you do not need to memorize the reference tables below; they exist for deviation and debugging.

If your host exposes a task/todo tool (e.g., TodoWrite), seed it with this checklist and tick phases as next moves past them:

- [ ] Phase 0: Literature grounding → lit_table.md, then Phase 0+ full-text fetch (MANDATORY — Phase 1 hard-gates on it)
- [ ] Phase 1: Bottleneck identification → phase1_output.json (routing: proceed | do_not_generate)
- [ ] Phase 2: Gap×pattern selection + candidate generation (ONE isolated context, TWO output files) → citation gate → coherence gate (dry-run trace, fresh context)
- [ ] Phase 3: Collision retrieval (signature@10mo + alias@48mo) → audit (5 checks) → [revise → merge → re-audit if falsification rewritten] | [abandon → 1 internal retry]
- [ ] Phase 4: skeleton → fill → assemble → implementability audit → validate → render → return 3 cards inline

Three outcomes per run: the rendered idea-card markdown returned inline (LaTeX + per-phase JSON left under $RUN_DIR), a do_not_generate.md (Phase 1 OOD), or a phase_3_failed.md (audit abandons twice). Never ask the user mid-flow — missing intake fields are inferred; revision, falsification re-audit, and the single internal retry all run without user re-invocation.

Invocation contract

No cd is required. scripts/run.py self-locates its skill root, so every orchestrator command can be invoked from ANY working directory by absolute script path: python3 "$SKILL_DIR/scripts/run.py" <subcommand> --out "$RUN_DIR/<phase>/" .... The legacy form cd "$SKILL_DIR" && python3 -m scripts.run <subcommand> ... works identically. Do NOT use relative script or --out paths — CWD is not stable across host-LLM Bash invocations, and the orchestrator rejects a relative --out outright.

Exit codes 10 and 11 are NOT errors — they are sentinel handshakes. When the orchestrator can't call an LLM itself (no NOVELTY_LLM_CLASSIFY_FAST_CMD), it writes a sentinel JSON describing what the host LLM should do, then exits rc=10 (intent / pattern-summary) or rc=11 (signature_terms). Read the sentinel ($RUN_DIR/<phase>/.<step>_pending), read the file at its rubric_file field (absolute path), produce the expected output, re-invoke per its re_invocation field. Do not stop on these codes. (The default Phase 0 flow below avoids the rc=10 intent sentinel entirely by passing --queries up front.)

Context discipline (read BEFORE running any LLM-driven phase)

A full run accumulates ~180-250k tokens of intermediate state. If the host LLM carries that in its own conversation context across phases, the Phase 1 / 2.2 / 4.fill calls routinely hit the backend request timeout ([API Error · Request timed out · Retrying...]) and the retry times out again. Apply ALL three rules on every run:

Rule 1 — Run every LLM-driven phase in an ISOLATED context. Phases 1 / 2 (2.1+2.2) / 2.3 / 3.2 / 3.3 / 4.fill / 4.1.5 each have file-path inputs and one JSON output; no phase needs the conversation that produced an earlier one. Use the FIRST isolation mechanism your harness supports:

• (a) Subprocess LLM — set NOVELTY_LLM_REASONING_LARGE_CMD / NOVELTY_LLM_CLASSIFY_FAST_CMD (see § Configuration); each phase runs as its own subprocess, fresh context by construction, on any harness.
• (b) Sub-agent tool (Claude Code Agent or equivalent) — spawn one per phase, passing ONLY the file paths the phase prompt lists — not conversation history, not file contents inline. The sub-agent reads from disk, Writes to disk, returns ≤ 250 words (output path + routing signal). Exception by design: Phase 2.1 and 2.2 run in ONE sub-agent writing both output files — both are generation-side; the adversarial separations (3.2 vs 3.3, 4.fill vs 4.1.5) must stay separate calls.
• (c) Manual context reset — run inline but clear/compact at the four points in Rule 3.

Whichever mechanism, the parent context stays ≤ ~30k tokens for the whole run because it never holds a phase's structured output.

Rule 2 — Write every phase artifact directly to disk; never paraphrase it into chat. Output convention: $RUN_DIR/<phase>/<phase>_output.json. Use your harness's file-write tool (Claude Code: Write) — no Bash heredocs (permission prompts + silent truncation), no echo, no pasting JSON into replies. Bound tool-result captures from large files to ≤ 4 KB (head -c 4000 / jq / sed); never Read a >10 KB intermediate dump into the parent context — the dump gets cached into every subsequent turn (this exact anti-pattern caused prior timeout runs).

Rule 3 — Compact between phases. Natural compact points: after Phase 0+, after Phase 1, after Phase 2, after Phase 3.2. Every phase re-reads its disk inputs, so compacting loses nothing. With /compact, use it there; Rule 1 mechanisms (a)/(b) achieve the same on their own.

Diagnostic for "Request timed out" mid-phase: inspect your harness's session transcript/log (Claude Code: ~/.claude/projects/<project-slug>/<session-id>.jsonl, look for isApiErrorMessage: true; other harnesses: their session-log equivalent); the prior tool call shows which prompt got too big. The fix is one of the three rules — usually Rule 1.

---

Phase reference

next prints each of these steps at the right moment with concrete paths; the tables below are the full contract for deviation/debugging.

Orchestrator entry points

Phase	Entry point (python3 "$SKILL_DIR/scripts/run.py" ..., any CWD)
navigator	next --dir "$RUN_DIR" [--query "..."]
Phase 0	phase0 --query "<user text>" --queries "q1|q2|q3|q4" --out $RUN_DIR/phase0/
user-ref registration (title-named anchor papers; BEFORE phase0_fulltext)	add_user_ref --out $RUN_DIR/phase0/ --title "<full title>" [--raw-match "<user phrasing>"] [--id <arxiv/DOI/URL>]
Phase 0+ full-text (mandatory, the moment lit_table.md lands)	phase0_fulltext --out $RUN_DIR/phase0/
Phase 1 anchor top-up (optional, when the #1 closest_adjacent fell outside the fulltext pool)	phase1_fulltext_topup --out $RUN_DIR/phase0/ --paper-id <anchor paper_id>
Phase 3.1 collision	phase3_collision --idea-json <canonical candidate> --out $RUN_DIR/phase3_collision/
Phase 3.3 merger	phase3_merge_revisions --phase2 <canonical candidate> --revisions <p3.3-patch> --critique <p3.2-report> --out $RUN_DIR/phase3_revise/
Phase 2.3 merger (same tool; only when coherence verdict=patched)	phase3_merge_revisions --phase2 <p2.2-output> --revisions <p2.3-output> --out $RUN_DIR/phase2_coherence/ --out-name refined_candidate.json
Phase 4 skeleton	phase4_skeleton --candidate <final_candidate-or-p2.2> --phase1 ... --phase2-select ... --phase3-critique ... [--phase3-revise ...] --phase0-dir $RUN_DIR/phase0/ [--collision ...] --out $RUN_DIR/phase4/
Phase 4 assemble	phase4_assemble --skeleton $RUN_DIR/phase4/phase4_skeleton.json --fill-map $RUN_DIR/phase4/fill_map.json --out $RUN_DIR/phase4/
Phase 4 render	phase4_render --expansion $RUN_DIR/phase4/phase4_expansion.json --out $RUN_DIR/phase4/
Validators	validate --phase2 ... [--phase3 ...] [--phase4 ...] [--phase4-impl ...]

The LLM-driven phases (1 / 2.1 / 2.2 / 2.3 / 3.2 / 3.3 / 4.fill / 4.1.5 / falsification re-audit) have no orchestrator subcommand (a cat prompt | llm wrapper would add fragility without determinism): read the prompt at references/system-prompts/<phase>.txt, gather the inputs listed at its top, Write the JSON described under Output: to $RUN_DIR/<phase>/<phase>_output.json. Run each under the Context discipline rules — Phase 4.fill is the largest output and the most timeout-prone; never in the parent context.

Phase 0 — Literature grounding

Phase 0 and 3.1 require real external retrieval via the bundled connector scripts (scripts/search_*.py) — never WebSearch or ad-hoc fetch (downstream phases reject unstructured output). Gate sentinel: .lit_grounding_mode = real vs connector_failure (halt with diagnostic; --allow-webfallback exists as a flagged, lower-confidence escape).

Default flow (skips one sentinel round-trip): BEFORE invoking phase0, read references/intent-recognition.md (Map mode) yourself and produce 4-6 search queries — including one ESCAPE-MECHANISM query phrased in solution vocabulary (recalls papers that already fixed the bottleneck and title themselves by their fix; problem-keyed queries miss exactly those). Also apply the OOD short-circuit (intake-routing.md triggers #1 Too-broad / #2 No-anchor → route to do_not_generate instead of retrieving). Then invoke with BOTH flags:

python3 "$SKILL_DIR/scripts/run.py" phase0 --query "<user's research question>" --queries "q1|q2|q3|q4" --out $RUN_DIR/phase0/

The rc=10 sentinel path still exists as fallback when --queries is omitted. The orchestrator: asserts a sane clock; probes 4 connectors and retrieves role-based (arxiv 0-6mo cap 10 / openalex 6-24mo cap 12 published-only / semanticscholar 6-24mo cap 13 published-only / openreview 0-6mo cap 10 in-review; ~40-45 papers; non-overlapping windows; SS-priority dedup); extracts URL/ID user-refs from the query into phase0/user_refs.json; emits .pattern_summary_pending for the host.

Retrieval takes 3-10 min (the openreview connector alone budgets 600s) — set your Bash timeout ≥ 600s or run it in the background.

Pattern tagging (host step): classify each lit_results.json paper per references/pattern-summary-rubric.md into 1-3 of the 15 patterns → write lit_table.md with columns paper_id | year_month | venue | title | ideation pattern tags | bottleneck this paper targets | open issue / unresolved gap | resolves_problem | retrieved_via. Pure classification — it does not need the large reasoning model: if your harness can route an isolated context to a cheaper/faster model tier or a lower reasoning effort, use that (this is the same tier NOVELTY_LLM_CLASSIFY_FAST_CMD names in § Configuration); otherwise just run it isolated on the host model.

Title-named user refs: if the user query names anchor papers by TITLE ("based on the LoRA paper" — anything the URL/ID regex can't catch), register each BEFORE phase0_fulltext via add_user_ref (entry-point table). It does a deterministic dedup-merge into user_refs.json — do NOT hand-edit that file (some harnesses' file-write tools refuse to overwrite a file that was never read, and a malformed edit silently drops the U fetch tier).

Phase 0+ full-text fetch — MANDATORY. The instant lit_table.md lands, run phase0_fulltext (entry-point table) before touching Phase 1. Pool = U (user refs, never capped) + T2 (top-10 published on-topic) + T3 (top-5 arxiv on-topic), ceiling 15 excluding U, method-first ordering, concurrent fetch (HTML path first, pymupdf PDF fallback; per-paper budget so one slow PDF can't stall the step). Output fulltext_cache.json keyed by paper_id ({tier, intro, method, source_used, warning}); fetch failures degrade to abstract + warning. Phase 1 hard-gates on this file (error: fulltext_not_fetched).

Phase 1 — Bottleneck identification

One isolated LLM call. Prompt: references/system-prompts/bottleneck_identify.txt. Inputs: user query + intake, phase0/lit_table.md, phase0/fulltext_cache.json (all-failed cache → continue with fulltext_degraded: true, abstract-level residue confidence), phase0/lit_results.json. Output phase1/phase1_output.json: intake (+_inferred_fields[] — missing fields are inferred, never asked), bottleneck_statement (≥2 paper_id cited inline), closest_adjacent[] ({paper_id, summary_and_residue}), what_phase_0_did_not_address[], state ∈ {proceed, do_not_generate}.

Routing: proceed (literature-groundable, no OOD trigger) or do_not_generate (too-broad / no-anchor OOD, <5 truly-relevant papers, genuinely blank space, or benchmark/system construction) → write do_not_generate.md with concrete remedial steps — terminal.

Phase 2 — Selection + generation (ONE isolated context, TWO outputs)

Run 2.1 and 2.2 back-to-back in one isolated context, writing BOTH output files (they are both generation-side; only adversarial pairs need separate calls):

2.1 — prompt references/system-prompts/ideate_select.txt; inputs phase1_output.json, references/ideation-patterns/overview.md (all 15 patterns' Definition / Operational signature / When to apply — selection at WHAT/WHEN level), references/ideation-patterns/companion-combos.md, lit_table.md. Pick the anchor gap (type-bound to intake.contribution_type), sample 1-3 sibling gaps randomly + coherence-filter (non-cohering siblings → deferred_gaps[]), judge each pattern per gap; record saturation (transparency, not a filter). Output phase2_select/phase2_select_output.json: selected_gaps[] (index 0 = anchor) + coherence_thread_type + pattern_saturation + deferred_gaps[]. Retry mode: when $RUN_DIR/attempt_1/ exists, the prompt's OPTIONAL retry input applies — the archived audit + selection become negative constraints.

2.2 — prompt references/system-prompts/ideate_generate.txt; inputs 2.1 output, phase1_output.json, lit_results.json, plus for each gap ONE picked sub-pattern card from references/ideation-sub-patterns/ (compare when_to_pick_this_one + differentiation_within_parent via its overview.md; then read the picked card's tactical_pattern + Step-by-Step). Output phase2_generate/phase2_generate_output.json — ONE candidate, 12 flat fields: title / hook / core_mechanism / core_mechanism_reasoning / core_mechanism_steps; gap_closure[] ({gap, main_pattern, sub_pattern: "C## (parent pattern name)", how_closed}, mirrors selected_gaps one-for-one); falsification_prediction (single paragraph: minimal experiment + metric-with-direction + ONE named load-bearing variable + negative control on that variable predicting the DOWNSTREAM outcome metric returns to baseline — non-tautological); compute_budget (user-relative, GPU-day line + API-dollar line when the campaign calls paid APIs; default intake envelope = 80GB-class GPUs, ≤8 concurrent, ≈150 GPU-days / 5 months, ~$10k API — overridable per user via IDEASPARK_DEFAULT_COMPUTE, see § Configuration); differentiation_from_lit[] (substantive deltas, not "different pattern"); almost_prior_paper_id + what_step_was_missed; signature_terms[] (own vocabulary — recent collision channel); alias_terms[] (other communities' names for the same mechanism, from parametric knowledge — multi-year alias collision channel). Both kill-switch fields (falsification_prediction, compute_budget) are locked from here on — see Phase 3 for the single audited exception.

Citation gate (deterministic, MANDATORY before Phase 3):

python3 "$SKILL_DIR/scripts/run.py" validate --phase2 $RUN_DIR/phase2_generate/phase2_generate_output.json

Any fail = a sub_pattern citation was guessed from the parent's gist, not read from overview.md. Fix against references/ideation-sub-patterns/overview.md (or regenerate 2.2 with the card open) and re-run until clean — the gate proves parent-consistency only; whether core_mechanism performs the cluster's actual tactic is Phase 3.2's recipe_application_check. (next runs this gate automatically.)

Coherence gate (2.3 — one isolated LLM call, MANDATORY after the citation gate, before 3.1): prompt references/system-prompts/coherence_trace.txt; inputs: the 2.2 candidate + the 2.1 spec; MUST be a FRESH context, never the 2.1+2.2 agent (the context that wrote a logic bug rubber-stamps it). It verifies internal procedural validity by EXECUTION, not review — four trace actions: formalize the dataflow (undefined symbols, missing producers, circular deps), numeric dry-run on one small concrete instance (magnitude/probability absurdities — logic bugs read fluently and only surface when computed), degenerate probes (empty/k=0/ties), and claim→step mapping (asserted properties nobody constructs). Verdict pass | patched; repairs are patch-only via the SAME merger (--out-name refined_candidate.json), scoped to making the written procedure sound (core_mechanism*, how_closed narrative, signature/alias terms when the repair changed what the mechanism is) — novelty surface, pattern bindings, and kill-switch fields are out of scope; unfixable-without-redesign findings go to unrepaired[] for the audit to weigh. Single pass, never abandons, and the 3.2 audit does NOT read its report (judges the repaired candidate blind). When refined_candidate.json exists it is the canonical candidate for every later phase (next wires this automatically). It validates that the algorithm survives on paper — NOT that it works empirically (falsification experiment) or is novel (audit).

Phase 3 — Quality gauntlet

3.1 collision (orchestrator, no LLM): entry-point table. TWO retrieval channels over all 4 connectors, merged into collision_hits.json with a per-hit collision_channel tag: signature — the candidate's signature_terms[] over a 10-month window (contemporaneous scoop risk); alias — the candidate's alias_terms[] (other communities' names for the same mechanism, produced from parametric knowledge at 2.2) over a 48-month window (renamed-ancestor risk — the "goal-conditioned success detector vs goal-image conditioned scorer" blind spot is lexical, not temporal, so widening the signature window alone cannot catch it). Missing signature_terms[] → rc=11 sentinel: produce BOTH term sets per intent-recognition.md Collision mode (terms 3-7 words each — long sentences break URL encoding), edit the candidate JSON, re-invoke. Missing only alias_terms[] → loud warning, alias channel skipped (add the field and re-run to close the blind spot). The audit-facing pool is relevance-truncated per channel (≤120 hits/channel by lexical overlap with the channel's own terms; zero-relevance BM25 noise dropped unconditionally; drops printed; untruncated pool preserved as collision_hits.full.json), so the audit can consume collision_hits.json in a few sequential Read chunks — no jq two-pass triage needed.

3.2 audit (one isolated LLM call): prompt references/system-prompts/critique.txt; inputs: candidate, 2.1 spec, lit_table.md, collision_hits.json, references/anti-patterns.md, and each cited sub-pattern card references/ideation-sub-patterns/<C##>.md (strip the leading code from sub_pattern; typically 1-3 cards, others NOT loaded). Five corpus-anchored checks:

Check	Question
gap_closure_reject_check	does the candidate match a documented Reject lesson in each cited sub-pattern card (## Tactical failure mode + ALL ### Reject lessons bullets)?
recipe_application_check	does core_mechanism actually perform the cited C## cluster's ## Tactical pattern signature move, or only the parent's generic idea (bypassed — the leading cause of incremental output)?
anti_pattern_check	if the SET of gap_closure[].main_pattern matches a reject-favored composition, is the required mitigation substantively delivered (artifact, not keyword)?
paper_pointed_threat	most specific subsuming/competing paper in lit_table ∪ collision_hits (both channels; alias-channel threats are NOT discounted for age); no_threat_found is valid — fabricating a generic threat is forbidden. Side output parametric_family_concern: a named un-retrieved mechanism family from parametric knowledge (family name + query vocabulary, never specific paper cites) — soft signal only, flows to Phase 4 reviewer_concerns as a "scoop-check X first" flag
falsification_structure_check	does falsification_prediction name the minimal experiment, the outcome metric + direction, ONE load-bearing variable, and a NON-tautological negative control targeting the downstream metric?

Verdict is two-layer. Hard floor (LLM cannot override) → abandon: triggered Reject lesson / unmitigatable anti-pattern / exact-mechanism collision. Soft judgment otherwise → advance (only trivial borderlines; concerns surface in Phase 4's reviewer_concerns) or revise with concrete revision_targets[] (scopes: tactical / sub_pattern / falsification). verdict_rationale must cite specific check findings. The audit judges only — it never modifies the candidate.

Routing on verdict:

• advance → Phase 4 reads the 2.2 candidate directly.

• revise → 3.3 (one isolated LLM call, prompt references/system-prompts/revise.txt): reads candidate + 2.1 spec + audit; emits patch-only applied_revisions[] — one entry per revision_target, ops replace / append_sentence / append_items / swap_sub_pattern / rewrite_falsification, never echoes the candidate, never re-judges the verdict. Then run the merger (entry-point table, WITH --critique) → writes phase3_revise/final_candidate.json + back-injects it into the patch file. Kill-switch fields are merger-refused with ONE audited exception: a scope=falsification target from falsification_structure_check is applied via the dedicated rewrite_falsification op (authorization verified against the audit report via --critique; same experiment/metric/claim, structure repaired; max one per run). When the merger prints falsification_rewritten, run the falsification re-audit (critique.txt § "Falsification re-audit mode": single check on final_candidate.json → phase3_critique/falsification_reaudit.json): advance → Phase 4; abandon → phase_3_failed.md. compute_budget has no revision route under any scope. No composition scope — gap-level changes route through the abandon-retry below, never through patches.

• abandon → one internal retry (the one-shot guarantee bars asking the user, not internal regeneration): if $RUN_DIR/.retry_used is absent, archive the attempt and regenerate —

  mkdir -p "$RUN_DIR/attempt_1" && \
  mv "$RUN_DIR/phase2_select" "$RUN_DIR/phase2_generate" "$RUN_DIR/phase2_coherence" "$RUN_DIR/phase3_collision" \
     "$RUN_DIR/phase3_critique" "$RUN_DIR/phase3_revise" "$RUN_DIR/attempt_1/" 2>/dev/null; \
  touch "$RUN_DIR/.retry_used"
  

  then re-run Phase 2 in retry mode (archived audit + selection = negative constraints), citation gate, 3.1, 3.2. Phase 0/1 artifacts are reused as-is. A second abandon → write phase_3_failed.md citing BOTH attempts' verdict_rationale + triggering checks + user-side options — terminal.

Phase 4 — Expansion + packaging

Five steps in order (next emits each with the correct flags for the advance vs revise path — on the revise path --candidate is final_candidate.json and --phase3-revise is passed; on advance it's the CANONICAL candidate (refined_candidate.json when 2.3 patched, else the 2.2 output) and the flag is omitted):

1. skeleton (orchestrator): populates every mechanical field — kill-switch echoes (byte-identical from the candidate), differentiation_from_lit venue_years, almost_prior_venue_year, why_prior_stopped[].paper_id/venue_year, domain_landscape (pattern_distribution + candidate_uses), literature_breakdown, reviewer_concerns_and_responses[].attack/severity/fields_changed_to_address (lifted from audit + patch), feasibility_validation.compute (bucketed against intake.compute) — and marks every prose field <TODO[path]: hint>.
2. fill (one isolated LLM call, prompt references/system-prompts/expand.txt): author ONLY the TODO paths as one flat {path: value} map → phase4/fill_map.json (~12 prose fields, ~8k tokens vs ~30 fields/~20k without the skeleton). No calendar projections; no experiment matrix / ablation plan / baseline table — the skill produces IDEA + falsifiability + feasibility judgment, not experimental engineering.
3. assemble (orchestrator): merges fill_map into skeleton → phase4_expansion.json; validates every path resolves to a real TODO; refuses kill-switch roots; warns on unfilled TODOs (expansion_completeness will fail them).
4. implementability audit (4.1.5, one isolated LLM call, default on): prompt references/system-prompts/implementability_audit.txt — fresh skeptical-engineer persona (separate from the 4.fill author) rewrites each method step into a buildable spec: enriched_steps[] (one per step, same ids/order, what_changes + what_to_do_en + what_to_do_zh) + underspecified_points[] ({step_id, hole, fill, severity: filled|open} — unfillable holes stay honest as open). Compute-agnostic by design (resource feasibility is 4.1's job); never adds/removes/renames steps; never carries kill-switch fields. Output phase4/phase4_implementability.json.
5. validate + render: run the validators (below), then phase4_render — templating only, no model call; auto-detects the sibling implementability file and merges enriched_steps by step_id into the rendered Method (deterministic; no-op when absent). Writes idea.std.zh.md (plain Chinese, domain-newcomer register) + idea.std.en.md (plain English) + idea.detail.en.md (rigorous English — the novelty + validity defense) + idea.std.{en,zh}.tex (auto-compiled to PDF when xelatex/tectonic is on PATH; skipped with a hint otherwise).

Final response: read all three markdown cards and return them inline under headings 中文版 / English / Reviewer version. Other phase outputs stay on disk for inspection, not echoed.

Validators

# advance path: --phase3 = phase3_critique_output.json; revise path: --phase3 = phase3_revise_output.json
# --phase2 = the CANONICAL candidate (refined_candidate.json when 2.3 patched, else the 2.2 output)
python3 "$SKILL_DIR/scripts/run.py" validate \
  --phase2 <canonical candidate file> \
  --phase3 <see comment> \
  --phase4 $RUN_DIR/phase4/phase4_expansion.json \
  --phase4-impl $RUN_DIR/phase4/phase4_implementability.json   # optional; enables implementability checks

Validator	Check	Severity
subpattern_citation_consistency	each gap_closure[].sub_pattern resolves to a real C## cluster in overview.md whose true parent == the cited main_pattern and whose parenthetical == that cluster's parent display name. Primary use: the Phase 2.2 citation gate; re-runs harmlessly here.	fail (hard)
kill_switch_integrity	falsification_prediction + compute_budget byte-identical along Phase 2.2 → [3.3 final_candidate →] 4. After an audited falsification rewrite (falsification_rewritten marker + matching applied rewrite_falsification entry — disagreement fails), the anchor for falsification_prediction re-bases at the 3.3 final_candidate (3.3 → 4 must match); compute_budget stays full-chain always.	fail (hard)
expansion_completeness	motivation (≥2 why_prior_stopped), method_flow.steps[] (each with linked_component + linked_falsification), feasibility_validation (5 sub-verdicts + overall), non-empty abstract_draft + core_claim + sub_claims[] — missing sections would render as silent blanks.	fail (hard)
implementability_completeness	enriched_steps[] one-per-step (same ids/order, EN+ZH), underspecified_points[] present ([] allowed), NO kill-switch field in the file.	fail (hard)
implementability_readability	std-register fields: no 占位/placeholder leak, no bare English jargon dropped into Chinese prose.	warn

Retry budget on fail (cap = 2). Fix only the named contract, re-validate; still failing after the 2nd retry → stop revising, render as-is, and append a short note listing the failing validators (a flagged-imperfect card beats a watchdog-killed run with zero output). Never "fix" kill_switch_integrity or subpattern_citation_consistency by editing a guarded field — surface them as the headline caveat instead.

Configuration

By default every model-driven phase runs on the host LLM. To route phases to a different backend (Gemini, open-weights, custom):

• NOVELTY_LLM_REASONING_LARGE_CMD — Phase 1 / 2.1 / 2.2 / 3.2 / 3.3 / 4.fill (needs ≥ 200k context, JSON output)
• NOVELTY_LLM_CLASSIFY_FAST_CMD — Phase 0 intent extraction + per-paper pattern tagging (smaller context, JSON output)

Each is a CLI taking a stdin prompt (<<SYSTEM>>...<<USER>>...) and emitting JSON on stdout. When unset (the default when running inside any host LLM), the orchestrator emits sentinel files and the host LLM handles those steps natively.

• IDEASPARK_DEFAULT_COMPUTE — optional standing compute profile for the user (free text, e.g. "8×H100 node, ~300 GPU-days, $50k API budget"). Put it in .env (auto-loaded); next surfaces it to Phase 1 as intake context. Precedence: compute stated in the user's query > this value > the factory default (80GB-class GPUs, ≤8 concurrent, ≈150 GPU-days / 5 months, ~$10k API campaign). Use this instead of editing the factory default — the default is the feasibility yardstick for users who state nothing.

Pipeline handoff (final step)

把最终 idea 卡片复制/汇总到 research_run/<课题slug>/stage1_ideas/ 下，并写入或追加 research_run/<课题slug>/stage1_ideas/idea_cards.md 索引；该阶段只产 1 张高质量卡。 脱离 research-pipeline 单独调用时，退当前工作目录。