comic-blind-comparison-review

Name: comic-blind-comparison-review
Author: wanshuiyin

44stars3forks

How to install comic-blind-comparison-review?

wanshuiyin/aris-movie-director/comic-blind-comparison-review

$npx -y skills add wanshuiyin/ARIS-Movie-Director --skill comic-blind-comparison-review

Installs into the current project.

Summary

Phase-1 comic-author step (post-final eval) — a DOUBLE-BLIND A/B of two FINAL whole comics: our cross-model-audited progressive render (the comic-author + comic-director output) vs a naive single-shot baseline. A single sealed coin-flip hides which is which; two cross-model revie

Files · 1

View on GitHub

SKILL.md

1	# comic-blind-comparison-review — Double-Blind A/B of Two Finished Comics (Phase 1, step S12, post-final)
2
3	The "vs baseline" capability the comic side otherwise lacks. The per-panel
4	[`comic-director`](../comic-director/SKILL.md) `panel_gate` and the cross-panel `assembly_gate` operate
5	during production on single units; this skill operates after production, on two complete works —
6	our cross-model-audited progressive comic (authored by the [`comic-author`](../comic-author/SKILL.md) suite,
7	baked by [`comic-director`](../comic-director/SKILL.md)) vs a naive single-prompt one-shot baseline — and
8	proves the spiral pipeline beats the one-shot, under a blind so neither reviewer can favor the home team.
9	It is a direct, faithful port of aris_movie's `blind-comparison-review` from video to a whole-comic A/B
10	(`outputs/aris_movie_progressive.mp4` vs `outputs/baseline_sd2_naive.mp4` → two rendered comic page sets), and
11	it is an evaluation skill, not an authoring one: it never edits a comic, it judges two finished ones.
12
13	> Cardinal lesson baked in as a gate, not prose: the order of operations IS the integrity guarantee.
14	> A single coin-flip seals `.ab_mapping.json` and `chmod 600`s it before any reviewer is invoked; the
15	> seal's mtime MUST precede both reviews; from seal-time until UNSEAL the orchestrator (Claude) may not read
16	> or quote the mapping; any home-team token in a reviewer prompt is a HARD-ABORT, not a warning. If the
17	> orchestrator reads the mapping before both reviews land, it may subconsciously phrase the synthesis to favor
18	> the home team — so the timestamp ordering is the audit trail, and "blinding integrity" is **REPORTED, never
19	> assumed**.
20
21	```text
22	progressive comic ┐ ┌─▶ comparison.md (Chinese deliverable)
23	(our audited) ├─▶ ⓪ SEAL .ab_mapping.json (1 coin-flip, │
24	baseline comic ┘ chmod 600, mtime BEFORE any reviewer) │
25	(naive one-shot) │ ▼ │
26	│ ① STAGE pages → A/ and B/ (sealed ids) │
27	│ ▼ │
28	│ ② CODEX blind review (xhigh, RO) ─────┤ (no home-team token; reads SHARED
29	│ ▼ │ blind spec = intent + ART_BIBLE only)
30	│ ③ GEMINI blind review (per-page + ─────┤
31	│ synthesis, auto-gemini-3) │
32	│ ▼ │
33	└─▶ ④ UNSEAL (verify seal_ts < BOTH reviews) ┘
34	▼
35	⑤ WRITE comparison.md + ⑥ WIKI A/B verdict nodes + edges
36	```
37
38	## Constants
39	- ARTIFACTS = two FINAL whole comics. `progressive` = our spiral output (`comic-director`'s baked frames +
40	the single-file viewer, or `comic.json` → render); `baseline` = a naive single-prompt one-shot comic of the
41	SAME story+style (no spiral, no per-panel gate). For the worked example the contract boundary is
42	[`examples/comic_m3_audit/comic.json`](../../examples/comic_m3_audit/comic.json) (the structured-IR artifact
43	under test) + [`examples/comic_m3_audit/ART_BIBLE.md`](../../examples/comic_m3_audit/ART_BIBLE.md) (the
44	SHARED blind spec both reviewers read).
45	- MAPPING_FILE = `outputs/.ab_mapping.json` — the seal, `chmod 600`. Written by ⓪ before any reviewer.
46	- REVIEWERS (two families, the load-bearing diversity) = Codex via a fresh `mcp__codex__codex` call,
47	`model_reasoning_effort: xhigh`, `sandbox: "read-only"` (no model pin — the call follows the local codex
48	CLI config; only the bake pins a model, via `run_comic.get_bake_plan()`) ‖ Gemini `auto-gemini-3` (per-page
49	`mcp__gemini__analyzeFile` + a text-only
50	`mcp__gemini-cli__ask-gemini` synthesis). Never downgrade the effort tier
51	([`reviewer-routing`](../../protocols/reviewer-routing.md)). Optionally fold in the

People who liked this also viewed

6 picks

Type

comic-blind-comparison-review

44stars3forks

View on GitHub

How to install comic-blind-comparison-review?

wanshuiyin/aris-movie-director/comic-blind-comparison-review

$npx -y skills add wanshuiyin/ARIS-Movie-Director --skill comic-blind-comparison-review

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# comic-blind-comparison-review — Double-Blind A/B of Two Finished Comics (Phase 1, step S12, post-final)
2
3	The "vs baseline" capability the comic side otherwise lacks. The per-panel
4	[`comic-director`](../comic-director/SKILL.md) `panel_gate` and the cross-panel `assembly_gate` operate
5	during production on single units; this skill operates after production, on two complete works —
6	our cross-model-audited progressive comic (authored by the [`comic-author`](../comic-author/SKILL.md) suite,
7	baked by [`comic-director`](../comic-director/SKILL.md)) vs a naive single-prompt one-shot baseline — and
8	proves the spiral pipeline beats the one-shot, under a blind so neither reviewer can favor the home team.
9	It is a direct, faithful port of aris_movie's `blind-comparison-review` from video to a whole-comic A/B
10	(`outputs/aris_movie_progressive.mp4` vs `outputs/baseline_sd2_naive.mp4` → two rendered comic page sets), and
11	it is an evaluation skill, not an authoring one: it never edits a comic, it judges two finished ones.
12
13	> Cardinal lesson baked in as a gate, not prose: the order of operations IS the integrity guarantee.
14	> A single coin-flip seals `.ab_mapping.json` and `chmod 600`s it before any reviewer is invoked; the
15	> seal's mtime MUST precede both reviews; from seal-time until UNSEAL the orchestrator (Claude) may not read
16	> or quote the mapping; any home-team token in a reviewer prompt is a HARD-ABORT, not a warning. If the
17	> orchestrator reads the mapping before both reviews land, it may subconsciously phrase the synthesis to favor
18	> the home team — so the timestamp ordering is the audit trail, and "blinding integrity" is **REPORTED, never
19	> assumed**.
20
21	```text
22	progressive comic ┐ ┌─▶ comparison.md (Chinese deliverable)
23	(our audited) ├─▶ ⓪ SEAL .ab_mapping.json (1 coin-flip, │
24	baseline comic ┘ chmod 600, mtime BEFORE any reviewer) │
25	(naive one-shot) │ ▼ │
26	│ ① STAGE pages → A/ and B/ (sealed ids) │
27	│ ▼ │
28	│ ② CODEX blind review (xhigh, RO) ─────┤ (no home-team token; reads SHARED
29	│ ▼ │ blind spec = intent + ART_BIBLE only)
30	│ ③ GEMINI blind review (per-page + ─────┤
31	│ synthesis, auto-gemini-3) │
32	│ ▼ │
33	└─▶ ④ UNSEAL (verify seal_ts < BOTH reviews) ┘
34	▼
35	⑤ WRITE comparison.md + ⑥ WIKI A/B verdict nodes + edges
36	```
37
38	## Constants
39	- ARTIFACTS = two FINAL whole comics. `progressive` = our spiral output (`comic-director`'s baked frames +
40	the single-file viewer, or `comic.json` → render); `baseline` = a naive single-prompt one-shot comic of the
41	SAME story+style (no spiral, no per-panel gate). For the worked example the contract boundary is
42	[`examples/comic_m3_audit/comic.json`](../../examples/comic_m3_audit/comic.json) (the structured-IR artifact
43	under test) + [`examples/comic_m3_audit/ART_BIBLE.md`](../../examples/comic_m3_audit/ART_BIBLE.md) (the
44	SHARED blind spec both reviewers read).
45	- MAPPING_FILE = `outputs/.ab_mapping.json` — the seal, `chmod 600`. Written by ⓪ before any reviewer.
46	- REVIEWERS (two families, the load-bearing diversity) = Codex via a fresh `mcp__codex__codex` call,
47	`model_reasoning_effort: xhigh`, `sandbox: "read-only"` (no model pin — the call follows the local codex
48	CLI config; only the bake pins a model, via `run_comic.get_bake_plan()`) ‖ Gemini `auto-gemini-3` (per-page
49	`mcp__gemini__analyzeFile` + a text-only
50	`mcp__gemini-cli__ask-gemini` synthesis). Never downgrade the effort tier
51	([`reviewer-routing`](../../protocols/reviewer-routing.md)). Optionally fold in the

People who liked this also viewed

6 picks

Type