evidence-draft

Name: evidence-draft
Author: WILLOSCAR

489stars39forks

How to install evidence-draft?

willoscar/research-units-pipeline-skills/evidence-draft

$npx -y skills add WILLOSCAR/research-units-pipeline-skills --skill evidence-draft

Installs into the current project.

Summary

Create per-subsection evidence packs (NO PROSE): claim candidates, concrete comparisons, evaluation protocol, limitations, plus citation-backed evidence snippets with provenance. Trigger: evidence draft, evidence pack, claim candidates, concrete comparisons, evidence snippets

Files · 1

View on GitHub

SKILL.md

1	# Evidence Draft
2
3	Build deterministic `outline/evidence_drafts.jsonl` packs from briefs + notes + optional evidence bindings.
4
5	Compatibility mode is active: this migration preserves the existing JSONL contract while moving evidence-quality policy, sparse-evidence routing, and evaluation-anchor rules into `references/` and `assets/`.
6
7	## Load Order
8
9	Always read:
10	- `references/overview.md`
11	- `references/evidence_quality_policy.md`
12
13	Read by task:
14	- `references/block_vs_downgrade.md` when deciding whether thin evidence should block drafting or only downgrade claim strength
15	- `references/evaluation_anchor_rules.md` when evaluation tokens, protocol context, or numeric claims are weak
16	- `references/examples_sparse_evidence.md` for evidence-thin pack calibration
17	- `references/source_text_hygiene.md` when paper self-narration or generic result wrappers are leaking into pack snippets / claim candidates
18
19	Machine-readable assets:
20	- `assets/evidence_pack_schema.json`
21	- `assets/evidence_policy.json`
22	- `assets/source_text_hygiene.json`
23
24	## Inputs
25
26	Required:
27	- `outline/subsection_briefs.jsonl`
28	- `papers/paper_notes.jsonl`
29	- `citations/ref.bib`
30
31	Optional but recommended:
32	- `papers/evidence_bank.jsonl`
33	- `outline/evidence_bindings.jsonl`
34
35	## Outputs
36
37	Keep the current output contract:
38	- `outline/evidence_drafts.jsonl`
39	- optional human-readable mirrors under `outline/evidence_drafts/`
40
41	## Script Boundary
42
43	Use `scripts/run.py` only for:
44	- deterministic joins across briefs / notes / evidence bank / bindings
45	- snippet extraction and provenance assembly
46	- policy-driven `blocking_missing` / `downgrade_signals` / `verify_fields` materialization
47	- pack validation and Markdown mirror generation
48
49	Do not treat `run.py` as the place for:
50	- filler bullets that make thin evidence look complete
51	- hidden sparse-evidence judgment that is not inspectable from `references/` / `assets/`
52	- reader-facing narrative prose
53
54	## Output Shape Rules
55
56	Keep these stable:
57	- preserve the existing top-level pack fields already used by downstream survey pipelines
58	- `claim_candidates` must remain snippet-derived
59	- `concrete_comparisons` must remain genuinely two-sided; if one cluster has no usable highlight, drop the card and surface thin evidence upstream instead of fabricating an A-vs-B contrast
60	- snippet sampling should stay cluster-aware: when a subsection has explicit clusters, evidence selection should avoid collapsing onto one route just because its abstracts contain louder result sentences
61	- sparse evidence should surface as explicit blockers / downgrade signals / verify fields, not filler bullets
62	- citation keys must remain constrained to `citations/ref.bib`
63
64	## Compatibility Notes
65
66	Current mode is reference-first with deterministic compatibility:
67	- `assets/evidence_policy.json` defines pack thresholds and sparse-evidence routing
68	- `assets/evidence_pack_schema.json` documents/validates the stable pack shape
69	- `scripts/run.py` still materializes the existing JSONL + Markdown outputs, but no longer pads sparse sections with generic caution prose
70
71	## Quick Start
72
73	- `uv run python .codex/skills/evidence-draft/scripts/run.py --workspace <workspace>`
74
75	## Execution Notes
76
77	When running in compatibility mode, `scripts/run.py` currently reads:
78	- `outline/subsection_briefs.jsonl`
79	- `papers/paper_notes.jsonl`
80	- `citations/ref.bib`
81	- optionally `papers/evidence_bank.jsonl` and `outline/evidence_bindings.jsonl`
82	- `assets/evidence_policy.json` and `assets/evidence_pack_schema.json`
83
84	## Script
85
86	### Quick Start
87
88	- `uv run python .codex/skills/evidence-draft/scripts/run.py --workspace <workspace>`
89
90	### All Options
91
92	- `--workspace <dir>`
93	- `--unit-id <id>`
94	- `--inputs <path1;path2>`
95	- `--outputs <path1;path2>`
96	- `--checkpoint <C*>`
97
98	### Examples
99
100	- `uv run python .codex/skills/evidence-draft/scripts/run.py --workspace <workspace>`
101
102	## Troubleshooting
103
104	- If packs look complete despite thin evidence, inspect `assets/evidence_policy.json` and `references/block_vs_downgrade.md` before changing Python.
105	- If evaluation bullets are generic, inspect `references/evaluation_anchor_rules.md` and the policy asset.
106	- If claims are strong but evidence is abstract/title-only, downgrade via `downgrade_signals` and `verify_fields` rather than adding narrative caveats.

People who liked this also viewed

6 picks

Type

evidence-draft

489stars39forks

View on GitHub

How to install evidence-draft?

willoscar/research-units-pipeline-skills/evidence-draft

$npx -y skills add WILLOSCAR/research-units-pipeline-skills --skill evidence-draft

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# Evidence Draft
2
3	Build deterministic `outline/evidence_drafts.jsonl` packs from briefs + notes + optional evidence bindings.
4
5	Compatibility mode is active: this migration preserves the existing JSONL contract while moving evidence-quality policy, sparse-evidence routing, and evaluation-anchor rules into `references/` and `assets/`.
6
7	## Load Order
8
9	Always read:
10	- `references/overview.md`
11	- `references/evidence_quality_policy.md`
12
13	Read by task:
14	- `references/block_vs_downgrade.md` when deciding whether thin evidence should block drafting or only downgrade claim strength
15	- `references/evaluation_anchor_rules.md` when evaluation tokens, protocol context, or numeric claims are weak
16	- `references/examples_sparse_evidence.md` for evidence-thin pack calibration
17	- `references/source_text_hygiene.md` when paper self-narration or generic result wrappers are leaking into pack snippets / claim candidates
18
19	Machine-readable assets:
20	- `assets/evidence_pack_schema.json`
21	- `assets/evidence_policy.json`
22	- `assets/source_text_hygiene.json`
23
24	## Inputs
25
26	Required:
27	- `outline/subsection_briefs.jsonl`
28	- `papers/paper_notes.jsonl`
29	- `citations/ref.bib`
30
31	Optional but recommended:
32	- `papers/evidence_bank.jsonl`
33	- `outline/evidence_bindings.jsonl`
34
35	## Outputs
36
37	Keep the current output contract:
38	- `outline/evidence_drafts.jsonl`
39	- optional human-readable mirrors under `outline/evidence_drafts/`
40
41	## Script Boundary
42
43	Use `scripts/run.py` only for:
44	- deterministic joins across briefs / notes / evidence bank / bindings
45	- snippet extraction and provenance assembly
46	- policy-driven `blocking_missing` / `downgrade_signals` / `verify_fields` materialization
47	- pack validation and Markdown mirror generation
48
49	Do not treat `run.py` as the place for:
50	- filler bullets that make thin evidence look complete
51	- hidden sparse-evidence judgment that is not inspectable from `references/` / `assets/`
52	- reader-facing narrative prose
53
54	## Output Shape Rules
55
56	Keep these stable:
57	- preserve the existing top-level pack fields already used by downstream survey pipelines
58	- `claim_candidates` must remain snippet-derived
59	- `concrete_comparisons` must remain genuinely two-sided; if one cluster has no usable highlight, drop the card and surface thin evidence upstream instead of fabricating an A-vs-B contrast
60	- snippet sampling should stay cluster-aware: when a subsection has explicit clusters, evidence selection should avoid collapsing onto one route just because its abstracts contain louder result sentences
61	- sparse evidence should surface as explicit blockers / downgrade signals / verify fields, not filler bullets
62	- citation keys must remain constrained to `citations/ref.bib`
63
64	## Compatibility Notes
65
66	Current mode is reference-first with deterministic compatibility:
67	- `assets/evidence_policy.json` defines pack thresholds and sparse-evidence routing
68	- `assets/evidence_pack_schema.json` documents/validates the stable pack shape
69	- `scripts/run.py` still materializes the existing JSONL + Markdown outputs, but no longer pads sparse sections with generic caution prose
70
71	## Quick Start
72
73	- `uv run python .codex/skills/evidence-draft/scripts/run.py --workspace <workspace>`
74
75	## Execution Notes
76
77	When running in compatibility mode, `scripts/run.py` currently reads:
78	- `outline/subsection_briefs.jsonl`
79	- `papers/paper_notes.jsonl`
80	- `citations/ref.bib`
81	- optionally `papers/evidence_bank.jsonl` and `outline/evidence_bindings.jsonl`
82	- `assets/evidence_policy.json` and `assets/evidence_pack_schema.json`
83
84	## Script
85
86	### Quick Start
87
88	- `uv run python .codex/skills/evidence-draft/scripts/run.py --workspace <workspace>`
89
90	### All Options
91
92	- `--workspace <dir>`
93	- `--unit-id <id>`
94	- `--inputs <path1;path2>`
95	- `--outputs <path1;path2>`
96	- `--checkpoint <C*>`
97
98	### Examples
99
100	- `uv run python .codex/skills/evidence-draft/scripts/run.py --workspace <workspace>`
101
102	## Troubleshooting
103
104	- If packs look complete despite thin evidence, inspect `assets/evidence_policy.json` and `references/block_vs_downgrade.md` before changing Python.
105	- If evaluation bullets are generic, inspect `references/evaluation_anchor_rules.md` and the policy asset.
106	- If claims are strong but evidence is abstract/title-only, downgrade via `downgrade_signals` and `verify_fields` rather than adding narrative caveats.

People who liked this also viewed

6 picks

Type