GBrain Skillpack JTBD Map
Product Frame
gbrain's skills turn a database and MCP server into a knowledge-work operating system. The 43 local SKILL.md files define how agents should ingest, query, enrich, publish, maintain, research, and operationalize memory.
The pattern is "thin harness, fat skills": deterministic TypeScript handles storage/search/sync, while skill files encode workflows, quality bars, routing, and judgment.
Skill Families
1. Always-On Brain Behavior
Skills: signal-detector, brain-ops.
Job: make every conversation improve the brain without forcing the user to remember capture hygiene.
Product spec:
- Detect entities, ideas, original thinking, and follow-up signals.
- Search brain before external APIs when prior context matters.
- Run read-enrich-write loops so responses create durable context.
2. Setup, Health, and Operations
Skills: install, setup, skillpack-check, smoke-test, maintain, testing, migrate, frontmatter-guard, citation-fixer.
Job: keep the brain running, coherent, and trustworthy.
Product spec:
- Verify environment, migrations, MCP registration, and skills.
- Audit stale pages, malformed frontmatter, missing citations, orphan links, dead links, and tag consistency.
- Run tests and smoke checks before relying on memory in high-value workflows.
3. Ingestion and Capture
Skills: ingest, idea-ingest, media-ingest, voice-note-ingest, meeting-ingestion, article-enrichment, archive-crawler, webhook-transforms, cold-start.
Job: convert messy real-world inputs into structured, searchable brain pages.
Product spec:
- Route by input type instead of asking the user to choose a workflow.
- Preserve exact phrasing for voice notes.
- Extract people, companies, concepts, timelines, quotes, and links.
- Enrich raw article/media dumps with summaries and why-it-matters analysis.
- Protect archives with explicit scan path configuration before crawling.
4. Research and Synthesis
Skills: book-mirror, strategic-reading, concept-synthesis, perplexity-research, academic-verify, data-research, cross-modal-review.
Job: make research personal, evidence-backed, and action-oriented.
Product spec:
- Compare new material against the user's existing brain.
- Separate what is new from what is already known.
- Preserve citations and methodology.
- Turn books/articles/cases into applied playbooks.
- Verify claims through source, method, data, and replication.
5. Retrieval and Knowledge Shaping
Skills: query, enrich, repo-architecture, functional-area-resolver, soul-audit.
Job: retrieve and improve knowledge without flattening it into generic summaries.
Product spec:
- Say "the brain does not have info" when evidence is absent.
- Route pages to correct architecture by subject, not input format.
- Enrich people/company pages with compiled truth and timelines.
- Resolve functional areas so content lands where future agents can find it.
6. Daily Workflow and Publishing
Skills: daily-task-manager, daily-task-prep, briefing, cron-scheduler, reports, publish, brain-pdf.
Job: turn memory into daily execution.
Product spec:
- Store tasks as searchable brain pages.
- Prepare for meetings with attendee and thread context.
- Generate briefings, reports, and PDFs from brain content.
- Publish selected pages as protected HTML.
- Schedule recurring maintenance and briefing flows.
7. Agent and Skill Meta-System
Skills: minion-orchestrator, ask-user, skill-creator, skillify.
Job: let the brain create or coordinate new workflows.
Product spec:
- Ask the user only when a missing decision blocks progress.
- Convert repeated workflows into skills.
- Coordinate sub-work with bounded roles and clear outputs.
- Preserve skill quality so workflow knowledge remains executable.
Persona Map
| Persona | Hires gbrain to... | Most Important Skills |
|---|---|---|
| Founder/operator | remember people, deals, meetings, decisions, and tasks | daily-task-prep, meeting-ingestion, enrich, briefing |
| Research-heavy builder | connect books, articles, and web research to existing context | book-mirror, strategic-reading, perplexity-research |
| AI coding operator | make repo/code lookup durable and symbol-aware | repo-architecture, query, setup, sync via gstack |
| Personal knowledge worker | capture voice notes, ideas, and archives without filing chores | voice-note-ingest, idea-ingest, archive-crawler |
| Team/admin operator | keep shared memory healthy and secure | maintain, citation-fixer, skillpack-check, migrate |
Key Product Insight
gbrain does not win by being a better notes app. It wins by making memory operational. The skillpack says: when an agent knows something, it should know where that fact lives, how confident it is, how it connects, and when to write the next update.
Complete Skill Catalog
This pass comes from reading gbrain/README.md and all 43 gbrain/skills/**/SKILL.md files. The useful product lens is that gbrain's skillpack is the behavior layer around memory: when to read, write, cite, enrich, defer, ask, verify, publish, and evaluate.
Always-On Brain Behavior
| Skill / Path | Trigger / When Used | Inputs | Outputs / writes_to | Gates / Checks | JTBD / Product Interpretation |
|---|---|---|---|---|---|
signal-detector<br>gbrain/skills/signal-detector/SKILL.md | Every inbound message | User message, entity/idea mentions | Silent page updates; people/, companies/, concepts/ | Must not block main response; cheap parallel subagent | Make memory compound ambiently without user filing effort. |
brain-ops<br>gbrain/skills/brain-ops/SKILL.md | Any brain read/write/lookup/citation | Query, target page/entity, citations | Enriched pages; people/, companies/, deals/, concepts/, meetings/ | Brain-first lookup; source attribution; backlinks | Core read-enrich-write loop that turns DB into operational memory. |
Setup, Bootstrap, Migration
| Skill / Path | Trigger / When Used | Inputs | Outputs / writes_to | Gates / Checks | JTBD / Product Interpretation |
|---|---|---|---|---|---|
install<br>gbrain/skills/install/SKILL.md | Legacy install intent | None; redirects to setup | Deprecated pointer | Use setup instead | Backward-compatibility shim. |
setup<br>gbrain/skills/setup/SKILL.md | "set up gbrain", initialize brain | Deployment topology, DB/PGLite/Supabase choice, API keys | Working GBrain install, AGENTS injection, sync verification | gbrain doctor --json; live sync test; no Supabase anon key | Reduce time-to-working-memory from days to minutes. |
cold-start<br>gbrain/skills/cold-start/SKILL.md | New brain asks "now what?" | Gmail, calendar, contacts, X archive, conversations, files | Bootstrapped pages; people/, companies/, meetings/, daily/, media/, conversations/, sources/ | Sample first 5 contacts; import only signal; ask before continuing | Convert empty install into useful personal brain. |
migrate<br>gbrain/skills/migrate/SKILL.md | "migrate from Obsidian/Notion/etc." | Existing vault/export/CSV/JSON/Roam | Migration report, pages, links/tags | Sample 5-10 files before bulk; round-trip verify; health check | Steal switching cost from notes apps. |
soul-audit<br>gbrain/skills/soul-audit/SKILL.md | Customize agent identity/profile | Interview answers | SOUL.md, USER.md, ACCESS_POLICY.md, HEARTBEAT.md | 6-phase interactive audit | Personalize the agent's operating identity and permissions. |
Ingestion And Capture
| Skill / Path | Trigger / When Used | Inputs | Outputs / writes_to | Gates / Checks | JTBD / Product Interpretation |
|---|---|---|---|---|---|
ingest<br>gbrain/skills/ingest/SKILL.md | "ingest/save/process this" | Arbitrary content | Routed pages; people/, companies/, concepts/, meetings/, sources/ | Detect type, delegate to specialized skill | One front door for messy inputs. |
idea-ingest<br>gbrain/skills/idea-ingest/SKILL.md | User shares URL/link/idea | Link, article, tweet, idea text | Pages; people/, concepts/, sources/ | Fetch, analyze, author page, cross-link | Turn external stimuli into durable thinking material. |
media-ingest<br>gbrain/skills/media-ingest/SKILL.md | Video, podcast, PDF, book, screenshot, repo | Media file/link/repo | Pages; concepts/, people/, companies/, sources/ | Transcript/extract entities/backlinks | Make non-text media searchable and relational. |
voice-note-ingest<br>gbrain/skills/voice-note-ingest/SKILL.md | Voice/audio note | Audio + transcript | Pages; voice-notes/, originals/, concepts/, people/, companies/, ideas/, personal/ | Preserve exact phrasing; store audio; cross-link entities | Capture raw original thought before it gets polished away. |
meeting-ingestion<br>gbrain/skills/meeting-ingestion/SKILL.md | Meeting transcript/notes | Transcript, attendees, companies | Pages; meetings/, people/, companies/ | Not complete until every entity is enriched | Turn meetings into relationship and timeline memory. |
article-enrichment<br>gbrain/skills/article-enrichment/SKILL.md | Raw article page needs structure | Article dump | Structured page; media/articles/ | Must include summary, quotes, insights, why-it-matters, see-also, raw source | Transform dumps into useful, quotable knowledge. |
archive-crawler<br>gbrain/skills/archive-crawler/SKILL.md | Crawl Dropbox/B2/local archive | Explicit archive paths | Pages; originals/, personal/, ideas/ | Refuses without archive-crawler.scan_paths; gold filter; manifest | Mine old personal archives safely for high-signal context. |
webhook-transforms<br>gbrain/skills/webhook-transforms/SKILL.md | Set up/process webhook | JSON event payload | Brain page/timeline update | Sanitize input; preserve raw payload; dead-letter on failure | Let external systems continuously feed the brain. |
Enrichment, Filing, Integrity
| Skill / Path | Trigger / When Used | Inputs | Outputs / writes_to | Gates / Checks | JTBD / Product Interpretation |
|---|---|---|---|---|---|
enrich<br>gbrain/skills/enrich/SKILL.md | Entity mentioned or page needs update | Person/company/entity | Pages; people/, companies/ | Tiered enrichment; compiled truth; backlinks/timeline | Make entity pages trustworthy and progressively richer. |
repo-architecture<br>gbrain/skills/repo-architecture/SKILL.md | "where does this go?" | Candidate content/page | Filing advice | File by primary subject, not format/source | Preserve future retrievability through consistent structure. |
frontmatter-guard<br>gbrain/skills/frontmatter-guard/SKILL.md | Validate/fix frontmatter | Brain page markdown | Repaired page metadata | YAML parse, slug, null bytes, delimiter checks | Stop malformed pages before they poison the index. |
citation-fixer<br>gbrain/skills/citation-fixer/SKILL.md | Citation audit/fix | Brain pages | Citation audit report, fixed pages | Inline source format; broken URL resolution | Protect credibility of memory claims. |
maintain<br>gbrain/skills/maintain/SKILL.md | Brain health, orphans, stale pages, dream cycle | Brain corpus | Health fixes, synthesized pages/patterns | Backlinks, citations, stale info, filing validation, graph backfill | Keep the brain from rotting as it grows. |
Query, Briefing, Publishing
| Skill / Path | Trigger / When Used | Inputs | Outputs / writes_to | Gates / Checks | JTBD / Product Interpretation |
|---|---|---|---|---|---|
query<br>gbrain/skills/query/SKILL.md | "what do we know", lookup/search | User question | Cited answer | Brain-only evidence; cite gaps/conflicts/stale info | Give agents memory-backed answers without hallucination. |
briefing<br>gbrain/skills/briefing/SKILL.md | Daily/morning briefing | Date, calendar, active deals/tasks | Daily briefing | Cite pages, timelines, recent changes | Turn stored memory into daily situational awareness. |
daily-task-prep<br>gbrain/skills/daily-task-prep/SKILL.md | Morning prep/day prep | Calendar, tasks, open threads | Actionable prep | Load attendee context and yesterday's threads | Make memory drive the day's work. |
daily-task-manager<br>gbrain/skills/daily-task-manager/SKILL.md | Add/complete/defer task | Task text/status | Updated task page/timeline | Priority lifecycle P0-P3 | Store tasks as searchable memory, not ephemeral todo text. |
reports<br>gbrain/skills/reports/SKILL.md | Save/load latest report | Report content/category | reports/{category}/{timestamp}.md | Timestamped retrieval by keyword | Make recurring outputs durable and queryable. |
publish<br>gbrain/skills/publish/SKILL.md | Share/publish page | Brain page slug | Protected HTML/share artifact | Zero LLM calls; password option | Let selected memory leave the brain safely. |
brain-pdf<br>gbrain/skills/brain-pdf/SKILL.md | Convert brain page to PDF | Brain page slug | PDF rendering | Strip frontmatter; sanitize emoji; use gstack make-pdf | Make brain pages presentation-grade without changing source. |
Research And Synthesis
| Skill / Path | Trigger / When Used | Inputs | Outputs / writes_to | Gates / Checks | JTBD / Product Interpretation |
|---|---|---|---|---|---|
book-mirror<br>gbrain/skills/book-mirror/SKILL.md | Personalized book analysis | EPUB/PDF + reader brain context | Personalized page; media/books/; optional PDF | Chapter extraction word counts; dense context pack; fact-check | Turn books into a mirror of the user's actual life. |
strategic-reading<br>gbrain/skills/strategic-reading/SKILL.md | Read through a specific problem lens | Source text + strategic problem | Playbook page; concepts/, projects/ | Cite every recommendation; direct quotes; short/medium/long plan | Convert reading into decisions and tactics. |
concept-synthesis<br>gbrain/skills/concept-synthesis/SKILL.md | Synthesize concepts / intellectual map | Concept pages | Tiered concept pages + concepts/README.md; concepts/ | Deduplicate, tier T1-T4, trace evolution | Turn scattered notes into an intellectual fingerprint. |
perplexity-research<br>gbrain/skills/perplexity-research/SKILL.md | Current web research / what changed | Topic + brain context | Research page; research/ | Separate new info from known context; citations | Brain-augmented web research with freshness delta. |
academic-verify<br>gbrain/skills/academic-verify/SKILL.md | Verify study/claim/citation | Research claim | Verdict page; concepts/ | Publication -> method -> raw data -> replication; verdict taxonomy | Add epistemic rigor to quoted studies. |
data-research<br>gbrain/skills/data-research/SKILL.md | Build tracker/extract data | YAML recipe, emails/docs/sources | Canonical trackers, raw archive | Deduplicate, archive raw sources, structured extraction | Turn messy source streams into structured operating data. |
cross-modal-review<br>gbrain/skills/cross-modal-review/SKILL.md | Second opinion/adversarial review | Diff, plan, brain page, bulk op | Review findings | Invoke for high-risk/bulk/security/100+ LOC; user sovereignty | Add model diversity as a quality gate. |
Scheduling, Jobs, Runtime Ops
| Skill / Path | Trigger / When Used | Inputs | Outputs / writes_to | Gates / Checks | JTBD / Product Interpretation |
|---|---|---|---|---|---|
cron-scheduler<br>gbrain/skills/cron-scheduler/SKILL.md | Schedule/quiet hours/jobs | Job name, cron, delivery policy | Saved job config | Staggering, collision prevention, quiet hours | Make autonomous memory work predictable and non-annoying. |
minion-orchestrator<br>gbrain/skills/minion-orchestrator/SKILL.md | Background job/subagent/fan-out | Shell job or agent task | Durable queue job/progress | Observable, steerable, pausable, replayable | Give agents durable async labor. |
skillpack-check<br>gbrain/skills/skillpack-check/SKILL.md | GBrain health / skillpack check | Local install state | JSON health report | Wraps doctor + migration list; exit code gates | One-glance operational health for host agents. |
smoke-test<br>gbrain/skills/smoke-test/SKILL.md | Restart/container health check | Environment/services | Pass/fail/autofix log | Critical service checks; known auto-fixes; exit code failures | Verify the brain survived runtime churn. |
testing<br>gbrain/skills/testing/SKILL.md | Run tests / validate skills | Repo/test suite/skills | Skill validation or regression report | Conformance, manifest, resolver, unit/eval/integration tiers | Make skill quality and system regressions observable. |
Skill Meta-System And Routing
| Skill / Path | Trigger / When Used | Inputs | Outputs / writes_to | Gates / Checks | JTBD / Product Interpretation |
|---|---|---|---|---|---|
ask-user<br>gbrain/skills/ask-user/SKILL.md | Choice gate / ambiguous decision | 2-3 choices + context | User decision prompt | Stop after asking; no preemptive action | Preserve human control at high-cost forks. |
skill-creator<br>gbrain/skills/skill-creator/SKILL.md | Create/improve skill | Skill idea | New skills/{name}/SKILL.md, manifest, resolver | MECE check; required sections | Let the system manufacture new procedures. |
skillify<br>gbrain/skills/skillify/SKILL.md | Make workflow "proper" | Raw feature/workflow | Skill tree, eval receipt, audit verdict | Required items, resolver, tests, cross-modal receipt | Productize ad hoc behavior into tested agent capability. |
functional-area-resolver<br>gbrain/skills/functional-area-resolver/SKILL.md | Compress AGENTS/RESOLVER | Routing file | Compressed functional-area dispatcher | A/B eval evidence; dispatcher clause required | Reduce context cost without losing routing accuracy. |
Deeper Product Insights
- gbrain's product is not the database; it is the behavioral layer around memory.
- The strongest design pattern is source of truth plus operational index: Markdown/pages remain legible while Postgres/PGLite/MCP make them executable by agents.
- The skillpack is organized around lifecycle, not content type. Capture, enrich, query, synthesize, publish, maintain, and evaluate are separate jobs.
- gbrain treats trust as a first-class product surface. Citation-fixer, academic-verify, frontmatter-guard, testing, smoke-test, and cross-modal-review are credibility infrastructure.
- The meta-skills are the platform move.
skillify,skill-creator,functional-area-resolver, andminion-orchestratorlet gbrain improve its own operating surface, not just store more information.