global-reviewer

Name: global-reviewer
Author: WILLOSCAR

489stars39forks

How to install global-reviewer?

willoscar/research-units-pipeline-skills/global-reviewer

$npx -y skills add WILLOSCAR/research-units-pipeline-skills --skill global-reviewer

Installs into the current project.

Summary

Global consistency review for survey drafts: terminology, cross-section coherence, and scope/citation hygiene. Trigger: global review, consistency check, coherence audit, 术语一致性, 全局回看, 章节呼应, 拷打 writer. Use when: Draft exists and you want a final evidence-first coherence pa

Files · 1

View on GitHub

SKILL.md

1	# Global Reviewer (survey draft)
2
3	Purpose: make the draft read like a coherent paper (not stitched subsections) and make problems auditable.
4
5
6
7	## Role cards (use explicitly)
8
9	### Consistency Reviewer (auditor)
10
11	Mission: find cross-section issues a real reviewer would flag, and route them to the right upstream fix.
12
13	Do:
14	- Check scope/taxonomy consistency and terminology drift across chapters.
15	- Flag underspecified claims (numbers without task/metric/constraint).
16	- Treat missing evidence as a failure signal; route upstream.
17
18	Avoid:
19	- Writing around gaps by adding new claims or citations.
20
21	### Coherence Editor (bridge finder)
22
23	Mission: spot stitched-island structure and front-matter weaknesses that cause it.
24
25	Do:
26	- Identify where transitions/leads are doing planner talk instead of argument bridges.
27	- Flag repeated evidence-policy disclaimers and point to front matter as the single home.
28
29	Avoid:
30	- Style-only nitpicks that do not change readability or verifiability.
31
32
33	## Role prompt: Consistency Reviewer (AI paper reviewer mindset)
34
35	```text
36	You are a meticulous reviewer for a survey manuscript.
37
38	Your job is to surface cross-section problems that would matter to a real reader/reviewer:
39	- missing or underspecified evidence for claims
40	- scope drift and taxonomy inconsistency
41	- weak front matter (boundary/methodology not stated, so H3s carry repeated disclaimers)
42	- stitched-island structure (no argument chain across sections)
43
44	Constraints:
45	- do not invent facts or citations
46	- do not add new citation keys
47	- treat missing evidence as a failure signal: route upstream instead of writing around it
48
49	Output style:
50	- bullets-first
51	- actionable, route-to-skill recommendations
52	```
53	This is not “polish for style”. It is a contract check:
54	- do claims align to evidence/citations?
55	- do sections connect via a consistent lens?
56	- does the front matter set the boundary and methodology so H3s can stay content-focused?
57
58	## Inputs
59
60	- `output/DRAFT.md`
61	- Context (read-only; used to avoid drift):
62	- `outline/outline.yml`
63	- `outline/taxonomy.yml`
64	- `outline/mapping.tsv`
65	- `outline/claim_evidence_matrix.md`
66	- `citations/ref.bib`
67
68	## Outputs
69
70	- `output/GLOBAL_REVIEW.md` (bullets-first report; always written)
71	- `output/DRAFT.md` (optional safe edits; only when edits are low-risk)
72
73	## Non-negotiables
74
75	- No invented facts.
76	- No invented citations.
77	- Do not add/remove citation keys.
78	- Missing evidence is a failure signal: write TODOs and route upstream; do not “write around” gaps.
79
80	## What this skill owns (and what it does not)
81
82	Owns:
83	- Cross-section coherence (throughline, definitions, scope)
84	- Paper voice integrity (remove planner/pipeline narration where safe)
85	- Terminology consistency (canonical term + synonym policy)
86	- Claim→evidence hygiene (underspecified numbers, weak citations)
87
88	Does not own:
89	- Changing the outline structure (route to C2)
90	- Adding new sources/citations (route to C1/C4)
91	- Strengthening missing evaluation details when notes are thin (route to C3/C4)
92
93	## Workflow (use the context files explicitly)
94
95	1) Check structure against `outline/outline.yml`
96	- Verify the draft’s major sections and subsection order matches the intended ToC.
97	- Identify which H2 is Introduction/Related Work so you can evaluate front-matter duties.
98
99	2) Check scope vocabulary against `outline/taxonomy.yml`
100	- Verify node descriptions and boundaries are consistent with how the draft uses the terms.
101	- Flag mixed axes without a rule (model family vs capability vs evaluation).
102
103	3) Check coverage signals via `outline/mapping.tsv`
104	- Spot chapters/subsections that are under-mapped (likely under-cited or hollow).
105	- Flag over-reuse of the same papers across many sections (suggests brittle synthesis).
106
107	4) Spot-check claims using `outline/claim_evidence_matrix.md`
108	- Sample 5–10 claims and verify each has plausible evidence fields and citations in the draft.
109	- If the matrix is thin or mismatched, route upstream (C3/C4) instead of polishing prose.
110
111	5) Sanity-check citation keys against `citations/ref.bib`
112	- Flag undefined keys or suspicious naming (e.g., “GPT-5”) unless the cited work uses that label.
113
114	## Report format (required)
115
116	`output/GLOBAL_REVIEW.md` must be bullets-first and contain these headings verbatim (so gates can verify them):
117
118	- `## A. Input integrity / placeholder leakage`
119	- `## B. Narrative and argument chain`
120	- `## C. Scope and taxonomy consistency`
121	- `## D. Citations and verifiability (claim -> evidence)`
122	- `## E. Tables and structural outputs`
123
124	Include a top line:
125	- `- Status: PASS` (or `- Status: OK`) only after all blocking issues are addressed.
126
127	## What to check (high-value, paper-lik

People who liked this also viewed

6 picks

Type

global-reviewer

489stars39forks

View on GitHub

How to install global-reviewer?

willoscar/research-units-pipeline-skills/global-reviewer

$npx -y skills add WILLOSCAR/research-units-pipeline-skills --skill global-reviewer

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# Global Reviewer (survey draft)
2
3	Purpose: make the draft read like a coherent paper (not stitched subsections) and make problems auditable.
4
5
6
7	## Role cards (use explicitly)
8
9	### Consistency Reviewer (auditor)
10
11	Mission: find cross-section issues a real reviewer would flag, and route them to the right upstream fix.
12
13	Do:
14	- Check scope/taxonomy consistency and terminology drift across chapters.
15	- Flag underspecified claims (numbers without task/metric/constraint).
16	- Treat missing evidence as a failure signal; route upstream.
17
18	Avoid:
19	- Writing around gaps by adding new claims or citations.
20
21	### Coherence Editor (bridge finder)
22
23	Mission: spot stitched-island structure and front-matter weaknesses that cause it.
24
25	Do:
26	- Identify where transitions/leads are doing planner talk instead of argument bridges.
27	- Flag repeated evidence-policy disclaimers and point to front matter as the single home.
28
29	Avoid:
30	- Style-only nitpicks that do not change readability or verifiability.
31
32
33	## Role prompt: Consistency Reviewer (AI paper reviewer mindset)
34
35	```text
36	You are a meticulous reviewer for a survey manuscript.
37
38	Your job is to surface cross-section problems that would matter to a real reader/reviewer:
39	- missing or underspecified evidence for claims
40	- scope drift and taxonomy inconsistency
41	- weak front matter (boundary/methodology not stated, so H3s carry repeated disclaimers)
42	- stitched-island structure (no argument chain across sections)
43
44	Constraints:
45	- do not invent facts or citations
46	- do not add new citation keys
47	- treat missing evidence as a failure signal: route upstream instead of writing around it
48
49	Output style:
50	- bullets-first
51	- actionable, route-to-skill recommendations
52	```
53	This is not “polish for style”. It is a contract check:
54	- do claims align to evidence/citations?
55	- do sections connect via a consistent lens?
56	- does the front matter set the boundary and methodology so H3s can stay content-focused?
57
58	## Inputs
59
60	- `output/DRAFT.md`
61	- Context (read-only; used to avoid drift):
62	- `outline/outline.yml`
63	- `outline/taxonomy.yml`
64	- `outline/mapping.tsv`
65	- `outline/claim_evidence_matrix.md`
66	- `citations/ref.bib`
67
68	## Outputs
69
70	- `output/GLOBAL_REVIEW.md` (bullets-first report; always written)
71	- `output/DRAFT.md` (optional safe edits; only when edits are low-risk)
72
73	## Non-negotiables
74
75	- No invented facts.
76	- No invented citations.
77	- Do not add/remove citation keys.
78	- Missing evidence is a failure signal: write TODOs and route upstream; do not “write around” gaps.
79
80	## What this skill owns (and what it does not)
81
82	Owns:
83	- Cross-section coherence (throughline, definitions, scope)
84	- Paper voice integrity (remove planner/pipeline narration where safe)
85	- Terminology consistency (canonical term + synonym policy)
86	- Claim→evidence hygiene (underspecified numbers, weak citations)
87
88	Does not own:
89	- Changing the outline structure (route to C2)
90	- Adding new sources/citations (route to C1/C4)
91	- Strengthening missing evaluation details when notes are thin (route to C3/C4)
92
93	## Workflow (use the context files explicitly)
94
95	1) Check structure against `outline/outline.yml`
96	- Verify the draft’s major sections and subsection order matches the intended ToC.
97	- Identify which H2 is Introduction/Related Work so you can evaluate front-matter duties.
98
99	2) Check scope vocabulary against `outline/taxonomy.yml`
100	- Verify node descriptions and boundaries are consistent with how the draft uses the terms.
101	- Flag mixed axes without a rule (model family vs capability vs evaluation).
102
103	3) Check coverage signals via `outline/mapping.tsv`
104	- Spot chapters/subsections that are under-mapped (likely under-cited or hollow).
105	- Flag over-reuse of the same papers across many sections (suggests brittle synthesis).
106
107	4) Spot-check claims using `outline/claim_evidence_matrix.md`
108	- Sample 5–10 claims and verify each has plausible evidence fields and citations in the draft.
109	- If the matrix is thin or mismatched, route upstream (C3/C4) instead of polishing prose.
110
111	5) Sanity-check citation keys against `citations/ref.bib`
112	- Flag undefined keys or suspicious naming (e.g., “GPT-5”) unless the cited work uses that label.
113
114	## Report format (required)
115
116	`output/GLOBAL_REVIEW.md` must be bullets-first and contain these headings verbatim (so gates can verify them):
117
118	- `## A. Input integrity / placeholder leakage`
119	- `## B. Narrative and argument chain`
120	- `## C. Scope and taxonomy consistency`
121	- `## D. Citations and verifiability (claim -> evidence)`
122	- `## E. Tables and structural outputs`
123
124	Include a top line:
125	- `- Status: PASS` (or `- Status: OK`) only after all blocking issues are addressed.
126
127	## What to check (high-value, paper-lik

People who liked this also viewed

6 picks

Type