一键导入
indexing-efficiency-worker
Implement hybrid event-driven indexing, incremental projection/embedding efficiency, and indexing reliability invariants.
用 Codex 或 Claude 帮你安装 复制这段 Prompt,粘贴到 Codex、Claude 或其他助手里,让它检查 Skill 页面并帮你完成安装。
菜单
Implement hybrid event-driven indexing, incremental projection/embedding efficiency, and indexing reliability invariants.
用 Codex 或 Claude 帮你安装 复制这段 Prompt,粘贴到 Codex、Claude 或其他助手里,让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Implement a bounded OpenBurnBar Windows parity milestone from reviewed VAL contracts with main-app composition, focused tests, baseline discipline, and Windows-host evidence preparation.
Runs a long-running mission using the Zenith continuous-improvement harness to dynamically orchestrate workers, testing, and stopping discipline.
Design consultation: understands your product, researches the landscape, proposes a complete design system (aesthetic, typography, color, layout, spacing, motion), and generates font+color preview pages. Creates DESIGN.md as your project's design source of truth. For existing sites, use /plan-design-review to infer the system instead. Use when asked to "design system", "brand guidelines", or "create DESIGN.md".
Use when asked about AI-agent spend, token usage, session history, workflow patterns, or cost investigations. Grounds all answers in OpenBurnBar/BurnBar local data via prompt context and MCP tools.
Run OpenBurnBar functional QA and write qa-results/report.md plus supporting artifacts.
Design finalization: generates production-quality Pretext-native HTML/CSS. Works with approved mockups from /design-shotgun, CEO plans from /plan-ceo-review, design review context from /plan-design-review, or from scratch with a user description. Text actually reflows, heights are computed, layouts are dynamic. 30KB overhead, zero deps. Smart API routing: picks the right Pretext patterns for each design type. Use when: "finalize this design", "turn this into HTML", "build me a page", "implement this design", or after any planning skill. Proactively suggest when user has approved a design or has a plan ready. (gstack) Voice triggers (speech-to-text aliases): "build the design", "code the mockup", "make it real".
| name | indexing-efficiency-worker |
| description | Implement hybrid event-driven indexing, incremental projection/embedding efficiency, and indexing reliability invariants. |
NOTE: Startup and cleanup are handled by worker-base. This skill defines the work procedure.
Use for features that modify projection jobs, indexing queues, chunk/embedding update logic, rebuild/re-embed flows, and cross-surface consistency between indexing and reporting.
None.
mission.md, AGENTS.md, .factory/library/architecture.md, and .factory/library/user-testing.md.fulfills, including edge-case invariants (stale jobs, lease recovery, pagination completeness).swift test --package-path OpenBurnBarCorexcodebuild test -only-testing for projection/index testsscripts/test-openburnbar-retrieval-evals.sh when retrieval/projection output changed{
"salientSummary": "Implemented event-driven dedupe and stale-version no-op handling in projection jobs, plus hash/version-based embedding skips. Added coverage for lease-recovery idempotency and remote reprojection re-enqueue after content change. Retrieval replay and scoped indexing tests pass.",
"whatWasImplemented": "Updated queue/job processing semantics so burst updates collapse to latest effective work, stale source-version jobs no-op without writes, and retry/reclaim paths remain idempotent. Added incremental chunk/embedding guards to avoid full rewrites when hashes and model versions are unchanged, and ensured remote conversation updates can enqueue new projection work after a prior completion.",
"whatWasLeftUndone": "",
"verification": {
"commandsRun": [
{
"command": "xcodebuild test -project OpenBurnBar.xcodeproj -scheme OpenBurnBar -destination \"platform=macOS,arch=arm64\" -only-testing:\"OpenBurnBarTests/ProjectionPipelineServiceTests\"",
"exitCode": 0,
"observation": "Projection job invariants and stale/no-op tests passed."
},
{
"command": "scripts/test-openburnbar-retrieval-evals.sh",
"exitCode": 0,
"observation": "Retrieval replay/golden suites remained green after incremental indexing changes."
},
{
"command": "swift test --package-path OpenBurnBarCore",
"exitCode": 0,
"observation": "Core package tests passed with new indexing fixtures."
}
],
"interactiveChecks": [
{
"action": "Manual: run repeated small-delta projection fixture and inspect reindex/reembed counters",
"observed": "No full rebuild trigger; only impacted chunks were rewritten."
}
]
},
"tests": {
"added": [
{
"file": "AgentLensTests/Active/ProjectionIncrementalContractsTests.swift",
"cases": [
{
"name": "test_staleSourceVersionJobNoOpsWithoutWrites",
"verifies": "Stale jobs complete without rewriting artifacts."
},
{
"name": "test_remoteUpdateCanReenqueueProjectionAfterCompletion",
"verifies": "Remote content updates can schedule a fresh projection pass."
}
]
}
]
},
"discoveredIssues": []
}
Every EndFeatureRun handoff MUST satisfy both of the following requirements to be considered valid:
The commitId field in handoffs MUST be a full machine-resolvable SHA-1 commit ID (40 hexadecimal characters). Short hashes or symbolic references (e.g., HEAD, main) are NOT acceptable.
Verification: Run git rev-parse <commitId> — it must resolve to the actual commit. If this command fails or returns a different hash, the handoff is non-compliant.
Why this matters: Short hashes and symbolic refs are ambiguous across clones and can become dangling references after rebases or force-pushes. Full SHAs are immutable and universally auditable.
followedProcedure=trueBefore setting followedProcedure: true in skillFeedback, you MUST demonstrate concrete evidence that you read the required context files listed in Phase 1.1 of worker-base. This is not self-certification — it requires showing what you actually read.
Required evidence format: In your verification.commandsRun array, include entries like:
{
"command": "head -n 3 mission.md && grep -n 'fulfills' AGENTS.md | head -n 5",
"exitCode": 0,
"observation": "Read mission.md goal section and AGENTS.md fulfills list during Phase 1.1 startup."
}
The evidence must demonstrate actual content was read (e.g., head, grep, cat, rg output showing file content), not merely that files exist (which test -f proves). For example, rg -n 'phase' AGENTS.md shows actual matching lines, while test -f AGENTS.md only proves the file exists.
When the final commit referenced by commitId is a non-implementation artifact commit (e.g., chore(...), docs(...), test(validation):..., or any commit that does not contain the feature's production/test code changes), the worker MUST also include the actual implementation commit SHA(s) in the handoff so that scrutiny reviewers can unambiguously map the implementation diff to the feature scope.
Implementation commits are commits whose diff contains the substantive code changes for the assigned feature (production code, test code, configuration changes that implement the feature behavior).
Artifact commits are commits whose diff is limited to validation synthesis, documentation updates, .factory/ state commits, or other non-implementation housekeeping.
How to provide implementation commit traceability:
git rev-parse HEAD after each implementation commit).commitId points to an artifact commit, include the implementation SHA(s) in handoff.whatWasImplemented with a clear reference, e.g.:
"whatWasImplemented": "Implemented X (implementation commit: abc1234...full SHA...). Final commit def... includes validation artifacts."
git show <implementation-sha> --stat to see the actual feature diff.Compliance: If your commitId references an artifact commit but no implementation commit SHA is included, followedProcedure MUST be set to false with a deviation documenting the omission.
If you deviated from the procedure: You MUST set followedProcedure: false and document every deviation in skillFeedback.deviations with:
step: Which phase/step was skipped or alteredwhatIDidInstead: What you actually didwhy: Blocking condition, better approach discovered, or unclear instructionAuditability: These requirements make every handoff independently verifiable:
git rev-parse and compared against git logworker-basecommitId is an artifact commit and whether implementation SHAs are provided