CtrlK
BlogDocsLog inGet started
Tessl Logo
SkillPublic

bulk-ingestion

End-to-end discipline for turning any large data source (audio libraries, email takeouts, document corpora, chat exports, API dumps) into brain pages at scale. The lifecycle spine: SCHEMA → ACCESS → TRIAL → EVALUATE → IMPROVE → CODIFY → TEST → SKILLIFY → BULK → MONITOR. State is tracked in a durable JSON manifest (see MANIFEST-PATTERN.md) so any crash, session boundary, or subagent fan-out resumes from ground truth instead of memory.

Quality

73%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide
SecuritybySnyk

Low

Low-risk findings worth noting

Fix and improve this skill with Tessl

tessl review fix ./plugin-variants/gbrain-daily/skills/bulk-ingestion/SKILL.md

The canonical home for this skill is bulk-ingestion in garrytan/gbrain

SKILL.md
Quality
Evals
Security

Quality

Content

77%Weight 40%Scale 1-5

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

A strong operational skill body: the ten-phase lifecycle is unambiguously sequenced with validation checkpoints, verification-over-trust rules, and error-recovery paths, and the guidance is concrete (real gbrain commands, file layouts, templates). Its two weaknesses are a missing companion file — the manifest pattern that Phases 2/6/9 depend on is referenced but not shipped in the bundle — and noticeable repetition of the trial-before-bulk rule across six sections.

Suggestions

Ship MANIFEST-PATTERN.md in the skill bundle (or inline its essentials: manifest.json item shape and status values, the idempotent-worker contract, and the sharding rule) — the body defers core Phase 2/9 detail to a file that is not present.

State the trial-before-bulk rule once (e.g., in the Contract) and reference it elsewhere; it is currently repeated in the conventions header, Contract, Phase 9, Quality Checklist, Anti-Patterns, and Related skills.

Show a minimal manifest.json example (items array with id/status plus a run-history entry) so Phase 2 and Phase 9 are executable without the missing companion file.

DimensionReasoningScore

Conciseness

The body is dense and prescriptive, delegates rather than re-derives ('Delegate to skills/skillify/SKILL.md ... Don't re-derive that checklist here'), and assumes competence throughout. The main trimmable redundancy is that the trial-before-bulk rule is restated roughly six times (conventions header, Contract, Phase 9 ladder, Quality Checklist, Anti-Patterns, Related skills). Efficient with minor instances of over-explanation — a 4, not 5 due to that repetition, not 3 since nearly every section is house-specific, non-obvious operational content.

4 / 5

Actionability

Concrete, executable guidance throughout: 'gbrain jobs submit shell --params '{"cmd": ...}'', 'gbrain agent run "Read skills/<pipeline-name>/SKILL.md and process the next 50 pending manifest items"', 'gbrain sync', 'gbrain check-backlinks check', a full YAML page template, the projects/<pipeline-name>/ artifact layout, and named pipeline responsibilities (fetchBatch(offset, limit), transformToPage(raw), ...). Not a 5 because the core artifact — the manifest.json structure — is never shown; its schema, the idempotent-worker contract, and sharding rules are all deferred to MANIFEST-PATTERN.md, which is not present in the bundle.

4 / 5

Workflow Clarity

Ten phases are explicitly sequenced with an ASCII spine, a mandated feedback loop ('Phases 3-5 loop until quality is satisfactory. Don't skip to bulk.'), gated ramping ('10 → 100 → 500 → full — with a quality check between rungs'), artifact-existence verification ('NEVER mark an item done without verifying its output artifact exists on disk'), a 12-item Quality Checklist before declaring done, and a dedicated 'When it fails' error-recovery section with exit codes. This matches the top anchor: clear sequence, explicit validation, feedback loops, and checklists — exactly what a batch-operation skill requires.

5 / 5

Progressive Disclosure

The body is well-sectioned and references are clearly signaled with read-timing ('read before Phase 2'), but the single most load-bearing reference — [MANIFEST-PATTERN.md](MANIFEST-PATTERN.md), cited ~6 times as the substrate for Phases 2, 6, and 9 — does not exist in the bundle (no references/, scripts/, or assets/ directories and no MANIFEST-PATTERN.md beside SKILL.md). Scoring against the actual bundle structure per the rubric guideline, the one-level-deep reference chain is broken for the skill's core mechanism. Not a 4 because a critical reference dangles; not a 2 because the body itself is well organized and nothing bulky is wrongly inlined.

3 / 5

Total

16

/

20

Passed

Description

58%Weight 40%Scale 1-5

Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.

A specific, well-scoped description with a clear 'what' (ten-stage lifecycle over a durable manifest) and good domain keyword coverage, but it omits the 'when': all trigger guidance ('Use when the user asks: "bulk ingest"...') sits in separate frontmatter fields instead of the description, leaving the description itself without use-when framing or the natural trigger verbs users say.

Suggestions

Append a 'Use when...' clause to the description itself, e.g. 'Use when the user asks to bulk ingest, bulk import, ingest all, or backfill a large set of items' — the trigger guidance currently lives only in the separate triggers/when_to_use frontmatter fields.

Work the natural trigger verbs ('bulk ingest', 'bulk import', 'ingest all', 'backfill') into the description text so the phrases users actually say appear there, not just the domain nouns (audio libraries, email takeouts, etc.).

Add one distinguishing phrase such as 'for enumerable sets of >~20 items, not one-off ingestion' to sharpen the boundary against single-item ingest skills.

DimensionReasoningScore

Specificity

The description enumerates many concrete actions: 'turning any large data source (audio libraries, email takeouts, document corpora, chat exports, API dumps) into brain pages at scale', the ten-stage 'lifecycle spine: SCHEMA → ACCESS → TRIAL → EVALUATE → IMPROVE → CODIFY → TEST → SKILLIFY → BULK → MONITOR', and 'State is tracked in a durable JSON manifest'. It is not a 5 because the stages are named as phase labels rather than fully fleshed-out capabilities, and not a 3 because it goes well beyond naming 1-2 actions.

4 / 5

Completeness

The 'what' is clear ('End-to-end discipline for turning any large data source ... into brain pages at scale' plus the lifecycle spine), but the description contains no 'Use when...' clause or equivalent trigger guidance — the 'when' is parked in the separate 'when_to_use' frontmatter field. Per the rubric guideline, a missing 'Use when...' clause caps completeness at 3; it cannot be a 4 because 'when' is wholly absent from the description text, not merely implicit.

3 / 5

Trigger Term Quality

Relevant domain nouns are present ('audio libraries, email takeouts, document corpora, chat exports, API dumps'), but the natural phrases a user would actually say — 'bulk ingest', 'bulk import', 'ingest all', 'backfill' — never appear in the description text itself; they live only in the separate frontmatter 'triggers' and 'when_to_use' fields. Some relevant keywords, missing the common variations — a 3, not 4 (the verbal triggers are absent) and not 2 (five concrete domain keywords are present).

3 / 5

Distinctiveness Conflict Risk

The description carves a clear niche — bulk, at-scale ingestion with 'a durable JSON manifest ... so any crash, session boundary, or subagent fan-out resumes from ground truth instead of memory' — that is distinct from generic processing skills. It is not a 5 because the description does not itself distinguish this skill from closely related siblings (single-item ingest, archive-crawler) that share the 'data source into brain pages' framing; minor overlap risk remains.

4 / 5

Total

14

/

20

Passed

Validation

87%

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation — 14 / 16 Passed

Validation for skill structure

CriteriaDescriptionResult

frontmatter_unknown_keys

Unknown frontmatter key(s) found; consider removing or moving to metadata

Warning

relative_links

Relative link issues: 5 missing, 12 suspicious

Warning

Total

14

/

16

Passed

Repository

garrytan/gbrain

Path

plugin-variants/gbrain-daily/skills/bulk-ingestion/SKILL.md

Reviewed

Table of Contents

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.