method-figure

Name: method-figure
Author: wanshuiyin

44stars3forks

How to install method-figure?

wanshuiyin/aris-movie-director/method-figure

$npx -y skills add wanshuiyin/ARIS-Movie-Director --skill method-figure

Installs into the current project.

Summary

Generate a publication-grade method / architecture / pipeline / workflow figure (a paper or README 'Figure 1') as an AUDITABLE object, not a one-shot prompt. A deterministic JSON blueprint LOCKS the content; an image model (gpt-image-2, baked by the agent via mcp__codex__codex —

Files · 1

View on GitHub

SKILL.md

1	# method-figure
2
3	Turn "draw our method figure" from a one-shot gamble into the **same audited spiral the framework uses for
4	comics**: a blueprint is the source of truth, the image model bakes the look, a cross-model panel + a
5	deterministic diff keep it honest, and the loop converges to a publication-grade figure that is
6	reproducible (re-run the blueprint) and auditable (a trace of every round).
7
8	Two things are simultaneously true: (a) gpt-image-2 CAN render a clean Figure-1 with legible labels when
9	conditioned on a labeled blueprint — do not assume it garbles text; (b) on a free prompt it DRIFTS
10	(renames phases, invents nodes, garbles a token, leaves pasted-looking floating labels). The blueprint +
11	blind-transcribe-then-hard-diff loop turns (a) into a reliable result and catches (b) every round.
12
13	```text
14	system description ─▶ ① BLUEPRINT (JSON content-lock) ── validate_blueprint.py
15	▼
16	② CONDITION (white-bg labeled SVG → PNG) + identity sheet (real chibi, optional) ── render_condition.py --png
17	▼
18	③ BAKE — agent: mcp__codex__codex(prompt+abs ref paths+out_path, workspace-write, gpt-5.5, config{xhigh}) → gpt-image-2 native PNG ── pickup_image.py --out-existing (sig+size+dims, mtime-bound, HARD-VETO struct/zlib/PIL/SVG, fail-closed)
19	▼
20	④ PANEL — Gemini ‖ Codex BLIND-transcribe → content_diff.py (observed ⊖ blueprint) → Claude structural sign-off
21	▼
22	⑤ agent reads the diff + the panel blockers → re-bake re-asserting the locked labels
23	▼
24	converged? ─ no ─▶ ③ (bounded: max_rounds → escalate to human)
25	│ yes
26	▼
27	⑥ APPROVE → figure.png + blueprint.json + trace.jsonl
28	```
29
30	## Constants
31	- GENERATOR = Codex `gpt-5.5`, `config: {model_reasoning_effort: xhigh, include_image_gen_tool: true}` → the native `image_generation`
32	tool (gpt-image-2). The `gpt-5.5` pin is a single hardcoded COMPAT DEFAULT in the bake sidecar payload
33	(`run_spiral.py` mirrors `run_comic.py`'s canonical bake plan; a config-driven model override is PLANNED, not
34	yet implemented). It pins the BAKE only — the panel's Codex reviewer is un-pinned (see PANEL below).
35	CRITICAL: image_gen is produced ONLY via `mcp__codex__codex` (the agent tool), NOT
36	`codex exec`. `codex exec` / over-specified / forbid-list prompts make Codex hand-draw a code fallback
37	(struct+zlib PNG or SVG/matplotlib) — visually indistinguishable for trivial shapes, useless for a real
38	method-figure. The working invocation is `mcp__codex__codex` with a dead-simple prompt +
39	`sandbox: "workspace-write"` (it must WRITE the out_path) + `model: "gpt-5.5"` +
40	`config: {model_reasoning_effort: "xhigh", include_image_gen_tool: true}` (the schema has NO top-level effort
41	param; `config{xhigh}` shorthand below ALWAYS expands to both these keys — without `include_image_gen_tool`
42	codex won't fire its native image tool, it falls back to descriptive text / an SVG renderer) + `cwd: <project>`.
43	Reference images are passed by absolute file path inside the prompt (the schema has NO `-i`); the output
44	path is a deterministic abs path in the prompt. Pick it up with `pickup_image.py --out-existing`
45	(verifies the EXPLICIT out_path: PNG sig + size + dims, `mtime >= request.created_at`) which HARD-VETOES
46	struct/zlib/PIL/`<svg>`/matplotlib markers in the agent transcript (fail-closed; there is no 'native sig
47	wins' override). Honesty caveat: as of Jun 2026 native headless persistence is unreliable, so this
48	fail-closed verifier — not any `sandbox` setting — is the first guard against a non-native bake. But the
49	HARD-VETO is a BEST-EFFORT denylist against the known codex-exec hand-draw fallback (struct/zlib/PIL/
50	SVG/mat

People who liked this also viewed

6 picks

Type

method-figure

44stars3forks

View on GitHub

How to install method-figure?

wanshuiyin/aris-movie-director/method-figure

$npx -y skills add wanshuiyin/ARIS-Movie-Director --skill method-figure

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# method-figure
2
3	Turn "draw our method figure" from a one-shot gamble into the **same audited spiral the framework uses for
4	comics**: a blueprint is the source of truth, the image model bakes the look, a cross-model panel + a
5	deterministic diff keep it honest, and the loop converges to a publication-grade figure that is
6	reproducible (re-run the blueprint) and auditable (a trace of every round).
7
8	Two things are simultaneously true: (a) gpt-image-2 CAN render a clean Figure-1 with legible labels when
9	conditioned on a labeled blueprint — do not assume it garbles text; (b) on a free prompt it DRIFTS
10	(renames phases, invents nodes, garbles a token, leaves pasted-looking floating labels). The blueprint +
11	blind-transcribe-then-hard-diff loop turns (a) into a reliable result and catches (b) every round.
12
13	```text
14	system description ─▶ ① BLUEPRINT (JSON content-lock) ── validate_blueprint.py
15	▼
16	② CONDITION (white-bg labeled SVG → PNG) + identity sheet (real chibi, optional) ── render_condition.py --png
17	▼
18	③ BAKE — agent: mcp__codex__codex(prompt+abs ref paths+out_path, workspace-write, gpt-5.5, config{xhigh}) → gpt-image-2 native PNG ── pickup_image.py --out-existing (sig+size+dims, mtime-bound, HARD-VETO struct/zlib/PIL/SVG, fail-closed)
19	▼
20	④ PANEL — Gemini ‖ Codex BLIND-transcribe → content_diff.py (observed ⊖ blueprint) → Claude structural sign-off
21	▼
22	⑤ agent reads the diff + the panel blockers → re-bake re-asserting the locked labels
23	▼
24	converged? ─ no ─▶ ③ (bounded: max_rounds → escalate to human)
25	│ yes
26	▼
27	⑥ APPROVE → figure.png + blueprint.json + trace.jsonl
28	```
29
30	## Constants
31	- GENERATOR = Codex `gpt-5.5`, `config: {model_reasoning_effort: xhigh, include_image_gen_tool: true}` → the native `image_generation`
32	tool (gpt-image-2). The `gpt-5.5` pin is a single hardcoded COMPAT DEFAULT in the bake sidecar payload
33	(`run_spiral.py` mirrors `run_comic.py`'s canonical bake plan; a config-driven model override is PLANNED, not
34	yet implemented). It pins the BAKE only — the panel's Codex reviewer is un-pinned (see PANEL below).
35	CRITICAL: image_gen is produced ONLY via `mcp__codex__codex` (the agent tool), NOT
36	`codex exec`. `codex exec` / over-specified / forbid-list prompts make Codex hand-draw a code fallback
37	(struct+zlib PNG or SVG/matplotlib) — visually indistinguishable for trivial shapes, useless for a real
38	method-figure. The working invocation is `mcp__codex__codex` with a dead-simple prompt +
39	`sandbox: "workspace-write"` (it must WRITE the out_path) + `model: "gpt-5.5"` +
40	`config: {model_reasoning_effort: "xhigh", include_image_gen_tool: true}` (the schema has NO top-level effort
41	param; `config{xhigh}` shorthand below ALWAYS expands to both these keys — without `include_image_gen_tool`
42	codex won't fire its native image tool, it falls back to descriptive text / an SVG renderer) + `cwd: <project>`.
43	Reference images are passed by absolute file path inside the prompt (the schema has NO `-i`); the output
44	path is a deterministic abs path in the prompt. Pick it up with `pickup_image.py --out-existing`
45	(verifies the EXPLICIT out_path: PNG sig + size + dims, `mtime >= request.created_at`) which HARD-VETOES
46	struct/zlib/PIL/`<svg>`/matplotlib markers in the agent transcript (fail-closed; there is no 'native sig
47	wins' override). Honesty caveat: as of Jun 2026 native headless persistence is unreliable, so this
48	fail-closed verifier — not any `sandbox` setting — is the first guard against a non-native bake. But the
49	HARD-VETO is a BEST-EFFORT denylist against the known codex-exec hand-draw fallback (struct/zlib/PIL/
50	SVG/mat

People who liked this also viewed

6 picks

Type