Skip to content

Lambenthan/empiricalwiki / i18n/en/skills/init

Bootstrap ΩmegaWiki from user sources plus optional discovery, then ingest the final paper set in parallel

FileKindSize
SKILL.mdskill md15.7 KB
references/parallel-ingest.mdmarkdown3.2 KB
references/planner-policy.mdmarkdown1.9 KB
references/prepare-and-discovery.mdmarkdown4.0 KB

references/planner-policy.md

View on GitHub
# /init Planner Policy

Use this reference when reading `.checkpoints/init-plan.json`, trimming the shortlist, or interpreting planner warnings and errors.

## Behavioral Policy

- In seeded mode, derive discovery and ranking cues from local paper titles/abstracts plus notes/web signals when `topic` is omitted.
- In bootstrap mode, if both `topic` and notes/web keywords are absent, surface a planner error instead of issuing an empty search.
- Favor relevance, freshness, connectivity, and survey coverage.
- Prefer one survey/overview when it improves coverage.
- In seeded mode with limited introduced capacity, freshness should dominate and older external non-survey papers should not pile up via citation advantage.
- In bootstrap mode or unusually roomy seeded cases, one older canonical anchor may be acceptable when it materially improves coverage.
- When Chinese note or web content is detected, keep a planner warning that extraction/ranking may be less reliable and treat provisional outputs as lower-confidence.
- If `SEMANTIC_SCHOLAR_API_KEY` is unset, continue anyway and preserve the slower public-rate-limit path as a planner warning.

## LLM Trim Expectations

- Read `.checkpoints/init-plan.json` and explicitly trim the over-picked shortlist before `fetch`.
- Do not skip the trim step even if the shortlist already looks reasonable.
- Emit a final selection artifact before `fetch` containing `shortlist_count`, `final_count`, and the final `candidate_id` list in shortlist order.
- If the final count is out of range, revise the selection before `fetch` unless a documented exception applies.

## Source Of Truth Boundary

- `tools/init_discovery.py` is the implementation authority for exact weights, thresholds, shortlist constants, and scoring math.
- `SKILL.md` and this reference describe orchestration and behavioral expectations only.
- Do not duplicate numeric planner constants here or override tool-owned policy during LLM reasoning.
add-product-e2e-evalpaperclipai/paperclipAdd or extend a Paperclip full-stack runner E2E workflow, fixture, matcher, or report evidence path for local or Daytona execution.Stars: 115kUpdated Claude CodeCodexVerifiedInstall
add-runner-evalpaperclipai/paperclipAdd or extend a Paperclip Runner protocol evaluation definition, roster, assertion, or report fixture with provenance and narrow validation.Stars: 115kUpdated Claude CodeCodexVerifiedInstall
diagnose-why-work-stoppedpaperclipai/paperclipCompatibility entry point for failed, stalled, looping, or over-recovered Paperclip work. Use fix-failed-paperclip-run for the full diagnosis, regression, repair, real verification, and green-PR workflow; honor an explicitly requested diagnosis-only scope.Stars: 115kUpdated Claude CodeCodexVerifiedInstall
fix-failed-paperclip-runpaperclipai/paperclipDiagnose and fix failed, stuck, slow, or incorrectly blocked Paperclip tasks and runs. Use with a task/run URL or failure report to trace the root cause, prove a regression red then green, check second-order effects, verify the real workflow, and deliver a PR with passing checks.Stars: 115kUpdated Claude CodeCodexVerifiedInstall