/contrib
Profile how top GitHub contributors work — ingest, build personas, and synthesise a capability model + learnable playbook.
$ npx -y skills add baodq97/tencentdb-agent-memory --agent claude-codeShips with tencentdb-agent-memory. Installing the plugin gets this command.
How it fires
How this command gets triggered: by you, by Claude, or both.
- Fires itselfClaude auto-loads it when your prompt matches the work.
- You can call itInvoke it directly when you want it.
- Slash command
/contrib
Context preview
What this command does when you run it.
Profile how top GitHub contributors work — ingest, build personas, and synthesise a capability model + learnable playbook.
Command definition
contrib.mddescription: Profile how top GitHub contributors work — ingest, build personas, and synthesise a capability model + learnable playbook.
/contrib
Contributor intelligence. Subjects are declared in `<global>/contributors/subjects.json`; all data lives in `<global>/contributors/index.db` (never the self-memory DB).
Subcommands
- `/contrib add <github_user> <owner/repo>` — add a subject.
- `/contrib ingest [id]` — fetch raw activity and extract 11-dimension atoms
(invokes the **contrib-ingest** skill). Omit id to ingest all subjects. Incremental by default (only activity since the last sync); pass `--full` to refetch everything.
- `/contrib build [id]` — consolidate atoms into a persona (invokes
**contrib-consolidate**).
- `/contrib persona <id>` — print one subject's dossier.
- `/contrib personas` — print all subjects' personas as JSON (for cross-engineer synthesis).
- `/contrib playbook <id>` — print the learnable playbook (invokes
**contrib-synthesize**).
- `/contrib compare <id>` — you vs a role model: a qualitative gap analysis of
the role model against your *existing* self-persona (`tmem persona`, built from your own history) — no need to ingest yourself from GitHub.
- `/contrib compare <id-a> <id-b>` — deterministic per-dimension table between two
*profiled contributors* (peer/team).
- `/contrib capabilities` — print the L4 capability model.
- `/contrib sync [id]` — embed atoms into the contributor vector index (FTS works
without this; vector recall needs it + the embed daemon).
- `/contrib search <query> [--subject <id>]` — keyword + vector recall over
atoms (FTS-only if the embed daemon is down).
- `/contrib trajectory <id>` — per-year cadence + commit-style evolution arc
(measures cadence/style, not PR LOC).
- `/contrib team add <teamId> <id...>` · `/contrib team capabilities <teamId>` —
group subjects and synthesise a team-level capability model.
Notes
- Requires an authenticated `gh` CLI. If missing, run `gh auth login`.
- Needs ≥2 subjects with personas before `capabilities`/L4 is meaningful (e.g. two
role models — you don't have to be one of them).
- "You vs role model" reuses your *existing* self-persona (`tmem persona`, built
from your own history) — you don't ingest your own GitHub.
Routing: subcommands that need judgment map to a skill — `ingest`/`build` → contrib-ingest/contrib-consolidate; `playbook`, `compare <id>` (single, you-vs- role-model), and the L4 narration of `capabilities` → contrib-synthesize. The deterministic subcommands — `add`, `persona`, `capabilities` (raw numbers), `compare <a> <b>` (two-contributor table), `trajectory`, `team`, `search` — call the matching `tmem contrib …` CLI directly.
Read more
description: Profile how top GitHub contributors work — ingest, build personas, and synthesise a capability model + learnable playbook.
/contrib
Contributor intelligence. Subjects are declared in `<global>/contributors/subjects.json`; all data lives in `<global>/contributors/index.db` (never the self-memory DB).
Subcommands
- `/contrib add <github_user> <owner/repo>` — add a subject.
- `/contrib ingest [id]` — fetch raw activity and extract 11-dimension atoms
(invokes the **contrib-ingest** skill). Omit id to ingest all subjects. Incremental by default (only activity since the last sync); pass `--full` to refetch everything.
- `/contrib build [id]` — consolidate atoms into a persona (invokes
**contrib-consolidate**).
- `/contrib persona <id>` — print one subject's dossier.
- `/contrib personas` — print all subjects' personas as JSON (for cross-engineer synthesis).
- `/contrib playbook <id>` — print the learnable playbook (invokes
**contrib-synthesize**).
- `/contrib compare <id>` — you vs a role model: a qualitative gap analysis of
the role model against your *existing* self-persona (`tmem persona`, built from your own history) — no need to ingest yourself from GitHub.
- `/contrib compare <id-a> <id-b>` — deterministic per-dimension table between two
*profiled contributors* (peer/team).
- `/contrib capabilities` — print the L4 capability model.
- `/contrib sync [id]` — embed atoms into the contributor vector index (FTS works
without this; vector recall needs it + the embed daemon).
- `/contrib search <query> [--subject <id>]` — keyword + vector recall over
atoms (FTS-only if the embed daemon is down).
- `/contrib trajectory <id>` — per-year cadence + commit-style evolution arc
(measures cadence/style, not PR LOC).
- `/contrib team add <teamId> <id...>` · `/contrib team capabilities <teamId>` —
group subjects and synthesise a team-level capability model.
Notes
- Requires an authenticated `gh` CLI. If missing, run `gh auth login`.
- Needs ≥2 subjects with personas before `capabilities`/L4 is meaningful (e.g. two
role models — you don't have to be one of them).
- "You vs role model" reuses your *existing* self-persona (`tmem persona`, built
from your own history) — you don't ingest your own GitHub.
Routing: subcommands that need judgment map to a skill — `ingest`/`build` → contrib-ingest/contrib-consolidate; `playbook`, `compare <id>` (single, you-vs- role-model), and the L4 narration of `capabilities` → contrib-synthesize. The deterministic subcommands — `add`, `persona`, `capabilities` (raw numbers), `compare <a> <b>` (two-contributor table), `trajectory`, `team`, `search` — call the matching `tmem contrib …` CLI directly.
Four-layer long-term memory (L0 Conversation → L1 Atom → L2 Scene → L3 Persona) for Claude Code, inspired by Tencent/TencentDB-Agent-Memory. Fully local — no external Gateway, no paid API, no Python.
Repo: baodq97/tencentdb-agent-memory