citation-injector

Name: citation-injector
Author: WILLOSCAR

35installs39forks

How to install citation-injector?

willoscar/research-units-pipeline-skills/citation-injector

$npx -y skills add WILLOSCAR/research-units-pipeline-skills --skill citation-injector

Installs into the current project.

Summary

Apply a citation-diversifier budget report by injecting *in-scope* citations into an existing draft (NO NEW FACTS), so the run passes the global unique-citation gate without citation dumps. Trigger: citation injector, apply citation budget, inject citations, add citations s

Files · 1

View on GitHub

SKILL.md

1	# Citation Injector (deterministic baseline edits; budget-as-constraints)
2
3	Purpose: make the pipeline converge when the draft is:
4	- locally citation-dense but globally under-cited (too few unique keys), or
5	- overly reusing the same citations across many subsections.
6
7	This skill is intentionally conservative and scriptable:
8	- the script edits `output/DRAFT.md` directly using the budget report as constraints
9	- injections stay evidence-neutral (NO NEW FACTS) and use only in-scope keys already listed for each H3
10
11	## Inputs
12
13	- `output/DRAFT.md`
14	- `output/CITATION_BUDGET_REPORT.md` (from `citation-diversifier`)
15	- `outline/outline.yml` (H3 id/title mapping)
16	- `citations/ref.bib` (must contain every injected key)
17
18	## Outputs
19
20	- `output/DRAFT.md` (updated in place)
21	- `output/CITATION_INJECTION_REPORT.md` (PASS/FAIL + what you changed)
22
23	## Non-negotiables (NO NEW FACTS)
24
25	- Only inject keys listed for that H3 in the budget report.
26	- Do not introduce new numbers, new benchmarks, or superiority claims.
27	- Do not add narration templates such as `This subsection discusses` or `Next, we examine`.
28	- Do not produce cite dumps like `[@a; @b; @c]` as the only citations in a paragraph.
29
30	## Paper-voice injection patterns (safe sentence shapes)
31
32	Use these as sentence intentions (paraphrase; do not copy verbatim).
33
34	1) Axis-anchored exemplars (preferred)
35	- `Systems such as X [@a] and Y [@b] instantiate <axis/design point>, whereas Z [@c] explores a contrasting point under a different protocol.`
36
37	2) Parenthetical grounding (short, low-risk)
38	- `The same design pressure appears in nearby systems (e.g., X [@a], Y [@b], Z [@c]).`
39
40	3) Cluster pointer + contrast hint
41	- `Representative implementations span both <cluster A> (X [@a], Y [@b]) and <cluster B> (Z [@c]), suggesting that the trade-off hinges on <lens>.`
42
43	4) Decision-lens pointer
44	- `For builders choosing between <A> and <B>, prior systems provide concrete instantiations on both sides (X [@a]; Y [@b]; Z [@c]).`
45
46	5) Evaluation-lens pointer (still evidence-neutral)
47	- `Across commonly used agent evaluations, systems such as X [@a] and Y [@b] illustrate how <lens> is operationalized, while Z [@c] highlights a different constraint.`
48
49	6) Contrast without list voice
50	- `While many works operationalize <topic> via <mechanism> (X [@a]; Y [@b]), others treat it as <alternative> (Z [@c]), which changes the failure modes discussed later.`
51
52	## Anti-patterns (high-signal “budget dump” voice)
53
54	Avoid these stems (they read like automated injection):
55	- `A few representative references include`
56	- `Notable lines of work include`
57	- `Concrete examples include`
58
59	If your draft contains these, rewrite them immediately using the patterns above (keep citation keys unchanged).
60
61	## Placement guidance
62
63	- Prefer inserting citations where the subsection already states a concrete contrast or decision lens.
64	- If you must add a new sentence/mini-paragraph, place it early (often after paragraph 1) so it reads as positioning, not as an afterthought.
65	- Keep injections subsection-specific: mention the subsection lens (H3 title / `contrast_hook`) so the same sentence cannot be copy-pasted into every H3.
66
67	## Workflow
68
69	1) Read the budget report (`output/CITATION_BUDGET_REPORT.md`)
70	- Treat `Global target (policy; blocking)` as the PASS line for the pipeline gate (derived from `queries.md:citation_target`; A150++ default: `recommended`).
71	- If `Gap: 0`, do nothing: write a short PASS report and move on.
72	- Otherwise, for each H3 with suggested keys, pick enough keys to close the gap to target:
73	- small gaps: 3-6 keys / H3
74	- A150++ gaps: often 6-12 keys / H3
75	Prefer keys that are unused globally and avoid repeating the same new keys across many H3s.
76
77	2) Inject in the right subsection
78	- Use `outline/outline.yml` to confirm H3 ordering and ensure the injected sentence lands inside the correct `###` subsection.
79
80	3) Inject with paper voice
81	- Prefer one short, axis-anchored sentence over a long enumerator sentence.
82	- Keep injections evidence-neutral (NO NEW FACTS) and avoid new numbers.
83	- Before you commit an injected key, confirm it exists in `citations/ref.bib`.
84
85	4) Write `output/CITATION_INJECTION_REPORT.md`
86	- Record which H3s you touched and which keys were added.
87	- Mark `- Status: PASS` only when the global target is met.
88
89	5) Verify
90	- Rerun the validator script (below) to recheck the global target.
91	- Then run `draft-polisher` to smooth any residual injection voice (citation keys must remain unchanged).
92
93	## Done criteria
94
95	- `output/CITATION_INJECTION_REPORT.md` exists and is `- Status: PASS`.
96	- `pipeline-auditor` no longer FAILs on “unique citations too low”.
97
98	##

People who liked this also viewed

6 picks

Type

citation-injector

35installs39forks

View on GitHub

How to install citation-injector?

willoscar/research-units-pipeline-skills/citation-injector

$npx -y skills add WILLOSCAR/research-units-pipeline-skills --skill citation-injector

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# Citation Injector (deterministic baseline edits; budget-as-constraints)
2
3	Purpose: make the pipeline converge when the draft is:
4	- locally citation-dense but globally under-cited (too few unique keys), or
5	- overly reusing the same citations across many subsections.
6
7	This skill is intentionally conservative and scriptable:
8	- the script edits `output/DRAFT.md` directly using the budget report as constraints
9	- injections stay evidence-neutral (NO NEW FACTS) and use only in-scope keys already listed for each H3
10
11	## Inputs
12
13	- `output/DRAFT.md`
14	- `output/CITATION_BUDGET_REPORT.md` (from `citation-diversifier`)
15	- `outline/outline.yml` (H3 id/title mapping)
16	- `citations/ref.bib` (must contain every injected key)
17
18	## Outputs
19
20	- `output/DRAFT.md` (updated in place)
21	- `output/CITATION_INJECTION_REPORT.md` (PASS/FAIL + what you changed)
22
23	## Non-negotiables (NO NEW FACTS)
24
25	- Only inject keys listed for that H3 in the budget report.
26	- Do not introduce new numbers, new benchmarks, or superiority claims.
27	- Do not add narration templates such as `This subsection discusses` or `Next, we examine`.
28	- Do not produce cite dumps like `[@a; @b; @c]` as the only citations in a paragraph.
29
30	## Paper-voice injection patterns (safe sentence shapes)
31
32	Use these as sentence intentions (paraphrase; do not copy verbatim).
33
34	1) Axis-anchored exemplars (preferred)
35	- `Systems such as X [@a] and Y [@b] instantiate <axis/design point>, whereas Z [@c] explores a contrasting point under a different protocol.`
36
37	2) Parenthetical grounding (short, low-risk)
38	- `The same design pressure appears in nearby systems (e.g., X [@a], Y [@b], Z [@c]).`
39
40	3) Cluster pointer + contrast hint
41	- `Representative implementations span both <cluster A> (X [@a], Y [@b]) and <cluster B> (Z [@c]), suggesting that the trade-off hinges on <lens>.`
42
43	4) Decision-lens pointer
44	- `For builders choosing between <A> and <B>, prior systems provide concrete instantiations on both sides (X [@a]; Y [@b]; Z [@c]).`
45
46	5) Evaluation-lens pointer (still evidence-neutral)
47	- `Across commonly used agent evaluations, systems such as X [@a] and Y [@b] illustrate how <lens> is operationalized, while Z [@c] highlights a different constraint.`
48
49	6) Contrast without list voice
50	- `While many works operationalize <topic> via <mechanism> (X [@a]; Y [@b]), others treat it as <alternative> (Z [@c]), which changes the failure modes discussed later.`
51
52	## Anti-patterns (high-signal “budget dump” voice)
53
54	Avoid these stems (they read like automated injection):
55	- `A few representative references include`
56	- `Notable lines of work include`
57	- `Concrete examples include`
58
59	If your draft contains these, rewrite them immediately using the patterns above (keep citation keys unchanged).
60
61	## Placement guidance
62
63	- Prefer inserting citations where the subsection already states a concrete contrast or decision lens.
64	- If you must add a new sentence/mini-paragraph, place it early (often after paragraph 1) so it reads as positioning, not as an afterthought.
65	- Keep injections subsection-specific: mention the subsection lens (H3 title / `contrast_hook`) so the same sentence cannot be copy-pasted into every H3.
66
67	## Workflow
68
69	1) Read the budget report (`output/CITATION_BUDGET_REPORT.md`)
70	- Treat `Global target (policy; blocking)` as the PASS line for the pipeline gate (derived from `queries.md:citation_target`; A150++ default: `recommended`).
71	- If `Gap: 0`, do nothing: write a short PASS report and move on.
72	- Otherwise, for each H3 with suggested keys, pick enough keys to close the gap to target:
73	- small gaps: 3-6 keys / H3
74	- A150++ gaps: often 6-12 keys / H3
75	Prefer keys that are unused globally and avoid repeating the same new keys across many H3s.
76
77	2) Inject in the right subsection
78	- Use `outline/outline.yml` to confirm H3 ordering and ensure the injected sentence lands inside the correct `###` subsection.
79
80	3) Inject with paper voice
81	- Prefer one short, axis-anchored sentence over a long enumerator sentence.
82	- Keep injections evidence-neutral (NO NEW FACTS) and avoid new numbers.
83	- Before you commit an injected key, confirm it exists in `citations/ref.bib`.
84
85	4) Write `output/CITATION_INJECTION_REPORT.md`
86	- Record which H3s you touched and which keys were added.
87	- Mark `- Status: PASS` only when the global target is met.
88
89	5) Verify
90	- Rerun the validator script (below) to recheck the global target.
91	- Then run `draft-polisher` to smooth any residual injection voice (citation keys must remain unchanged).
92
93	## Done criteria
94
95	- `output/CITATION_INJECTION_REPORT.md` exists and is `- Status: PASS`.
96	- `pipeline-auditor` no longer FAILs on “unique citations too low”.
97
98	##

People who liked this also viewed

6 picks

Type