research-stewardship

Name: research-stewardship
Author: pariyar07

18stars2forks

How to install research-stewardship?

pariyar07/ariadne/research-stewardship

$npx -y skills add pariyar07/ariadne --skill research-stewardship

Installs into the current project.

Summary

Use when auditing or repairing provenance, compilation coverage, stale synthesis, unresolved questions, circular corroboration, nested-boundary isolation, or legacy research drift within one named Markdown knowledge-vault research boundary.

Files · 1

View on GitHub

SKILL.md

1	# Ariadne Research Stewardship
2
3	Audit one named research boundary first. Repair only deterministic, meaning-preserving drift through a declared exact write set.
4
5	## Authorization And Modes
6
7	1. Require a current-turn named or confirmed research boundary in a multi-scope vault. A likely match is not authorization.
8	2. Read instructions from the vault root through every ancestor scope to the target. Read the boundary descriptor, canonical hubs, inquiry and synthesis entrypoints, and validator output progressively.
9	3. Choose `audit`, `repair`, or `legacy-adoption` explicitly.
10	4. In `audit` mode perform zero mutations: no reports in the vault, metadata normalization, link edits, queue items, or opportunistic fixes. Return findings and a proposed exact write set in chat.
11	5. In `repair` mode state the exact `allowed_write_set` before editing. Keep every mutation inside it. A newly discovered path is deferred until it is explicitly added.
12
13	The selected boundary does not authorize sibling scopes, parent/root navigation, or downstream destinations. Folder ancestry does not establish boundary membership.
14
15	## Audit Contract
16
17	Audit the declared boundary for:
18
19	- descriptor and hub consistency
20	- canonical downstream-to-upstream `derived_from` provenance
21	- raw sources marked `compiled` without a derived note
22	- source-only and needs-review coverage
23	- possible provenance cycles or circular corroboration
24	- evidence roles and generated or derivative evidence families
25	- stale synthesis, contradictions, supersession candidates, and unresolved questions
26	- exact versus explicit descendant rollup behavior
27	- nested child boundary isolation
28	- older pipeline drift and adoption readiness
29
30	Report semantic ambiguity; do not guess. A structurally valid boundary may still be semantically stale.
31
32	## Automatic Repair Allowlist
33
34	Automatic repair is limited to:
35
36	- path-qualifying a uniquely resolved link
37	- restoring a descriptor-declared hub edge
38	- linking an existing Base from its existing sibling Base index
39	- normalizing a schema value without changing its meaning
40
41	No other automatic repair is allowed. Each edit must be supported by existing canonical structure, remain inside the exact write set, preserve user content, and be idempotent: a second repair run produces no further mutations.
42
43	Always defer evidence-role classification, provenance-root merging or deletion, contradiction or supersession resolution, compilation versus source-only decisions, synthesis rewrites, promotion, moves, renames, merges, deletion, cross-scope rehoming, and any ambiguous link target.
44
45	## Nested Children And Legacy Adoption
46
47	Read a nested child boundary for context when needed, but do not repair it under the parent's contract. Before any child edit, read the child's descriptor plus all applicable ancestor and local instructions, then construct and declare a child-specific exact `allowed_write_set`. Do not inherit or reuse the parent's set. Parent conventions never overwrite a child's explicit contract. A parent `rollup` includes only the child descriptors listed in `rollup_boundaries`.
48
49	Treat older pipelines without a supported `research_schema` as legacy. Audit may inventory them, but adoption is opt-in and scope-local. In `legacy-adoption` mode map existing structure first, preserve stable paths, adopt before creating parallels, and add a descriptor only after mapping is unambiguous or confirmed. Apply only allowlisted repairs automatically; propose moves, renames, template changes, and synthesis changes for review.
50
51	## Validation And Report
52
53	After repair or opt-in adoption, run scoped research validation first:
54
55	```text
56	validate_vault.js "/path/to/vault" --scope "<scope_path>" --profile research
57	```
58
59	Then run whole-vault validation to detect collateral effects. Re-run stewardship in audit mode or compare hashes to prove idempotency. Scoped findings must remain a subset of whole-vault findings for the same state.
60
61	Report mode, target boundary, instructions read, findings, exact write set, repairs, deferrals, scoped validation, whole-vault validation, and second-run idempotency. Recommended cadence is weekly for active or automated-ingest research, monthly for stable research, and event-driven after large imports, major synthesis changes, or detected drift.

People who liked this also viewed

6 picks

Type

research-stewardship

18stars2forks

View on GitHub

How to install research-stewardship?

pariyar07/ariadne/research-stewardship

$npx -y skills add pariyar07/ariadne --skill research-stewardship

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# Ariadne Research Stewardship
2
3	Audit one named research boundary first. Repair only deterministic, meaning-preserving drift through a declared exact write set.
4
5	## Authorization And Modes
6
7	1. Require a current-turn named or confirmed research boundary in a multi-scope vault. A likely match is not authorization.
8	2. Read instructions from the vault root through every ancestor scope to the target. Read the boundary descriptor, canonical hubs, inquiry and synthesis entrypoints, and validator output progressively.
9	3. Choose `audit`, `repair`, or `legacy-adoption` explicitly.
10	4. In `audit` mode perform zero mutations: no reports in the vault, metadata normalization, link edits, queue items, or opportunistic fixes. Return findings and a proposed exact write set in chat.
11	5. In `repair` mode state the exact `allowed_write_set` before editing. Keep every mutation inside it. A newly discovered path is deferred until it is explicitly added.
12
13	The selected boundary does not authorize sibling scopes, parent/root navigation, or downstream destinations. Folder ancestry does not establish boundary membership.
14
15	## Audit Contract
16
17	Audit the declared boundary for:
18
19	- descriptor and hub consistency
20	- canonical downstream-to-upstream `derived_from` provenance
21	- raw sources marked `compiled` without a derived note
22	- source-only and needs-review coverage
23	- possible provenance cycles or circular corroboration
24	- evidence roles and generated or derivative evidence families
25	- stale synthesis, contradictions, supersession candidates, and unresolved questions
26	- exact versus explicit descendant rollup behavior
27	- nested child boundary isolation
28	- older pipeline drift and adoption readiness
29
30	Report semantic ambiguity; do not guess. A structurally valid boundary may still be semantically stale.
31
32	## Automatic Repair Allowlist
33
34	Automatic repair is limited to:
35
36	- path-qualifying a uniquely resolved link
37	- restoring a descriptor-declared hub edge
38	- linking an existing Base from its existing sibling Base index
39	- normalizing a schema value without changing its meaning
40
41	No other automatic repair is allowed. Each edit must be supported by existing canonical structure, remain inside the exact write set, preserve user content, and be idempotent: a second repair run produces no further mutations.
42
43	Always defer evidence-role classification, provenance-root merging or deletion, contradiction or supersession resolution, compilation versus source-only decisions, synthesis rewrites, promotion, moves, renames, merges, deletion, cross-scope rehoming, and any ambiguous link target.
44
45	## Nested Children And Legacy Adoption
46
47	Read a nested child boundary for context when needed, but do not repair it under the parent's contract. Before any child edit, read the child's descriptor plus all applicable ancestor and local instructions, then construct and declare a child-specific exact `allowed_write_set`. Do not inherit or reuse the parent's set. Parent conventions never overwrite a child's explicit contract. A parent `rollup` includes only the child descriptors listed in `rollup_boundaries`.
48
49	Treat older pipelines without a supported `research_schema` as legacy. Audit may inventory them, but adoption is opt-in and scope-local. In `legacy-adoption` mode map existing structure first, preserve stable paths, adopt before creating parallels, and add a descriptor only after mapping is unambiguous or confirmed. Apply only allowlisted repairs automatically; propose moves, renames, template changes, and synthesis changes for review.
50
51	## Validation And Report
52
53	After repair or opt-in adoption, run scoped research validation first:
54
55	```text
56	validate_vault.js "/path/to/vault" --scope "<scope_path>" --profile research
57	```
58
59	Then run whole-vault validation to detect collateral effects. Re-run stewardship in audit mode or compare hashes to prove idempotency. Scoped findings must remain a subset of whole-vault findings for the same state.
60
61	Report mode, target boundary, instructions read, findings, exact write set, repairs, deferrals, scoped validation, whole-vault validation, and second-run idempotency. Recommended cadence is weekly for active or automated-ingest research, monthly for stable research, and event-driven after large imports, major synthesis changes, or detected drift.

People who liked this also viewed

6 picks

Type