Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add commands/m3talux/superbrain/discovergit clone --depth 1 https://github.com/m3talux/superbrainWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00084 | $0.01020 |
| Opus 5 | $0.00042 | $0.00510 |
| Sonnet 5 | $0.00017 | $0.00204 |
| Haiku 4.5 | $0.00008 | $0.00102 |
Grade A, and why
discover scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 59 lines — stays where its author put it; the contents beside it link to each section on GitHub.
/superbrain:discover
Manual project discovery. Auto-discovery only fires on first session in a project with no existing note — this command bypasses that gate.
What it does
- Walks the project tree (bounded: 600 files, depth 5, vendored/build dirs skipped).
- Reads manifests +
CLAUDE.md+README+Dockerfile(each capped at 8 KB). - Invokes one detached
claude -p sonnet-4-6call with the gathered context. - Writes a structured project note covering: stack, architecture, top-level folders, key files, docs, conventions, open questions.
- If the project note already exists, appends a fresh
## SuperBrain discovery (YYYY-MM-DD)section at the bottom — never overwrites user content. If no note exists, creates one with fulldiscovered: truefrontmatter.
How to invoke
Step 1 — determine the target project directory:
- If an argument was passed to the slash command, treat it as an absolute path (or relative to the current working directory). Verify it exists and is a directory.
- Otherwise, default to
$CLAUDE_PROJECT_DIR(the project the current session opened in).
Step 2 — invoke the discoverer binary. The plugin classifies the target through the four-stage detection cascade (blocklist → strong-signal → workspace declaration → implicit umbrella) and handles single-project vs umbrella fan-out internally.
node "${CLAUDE_PLUGIN_ROOT}/dist/bin/sb-discover.js" --force "<project_dir>"
If the user explicitly asked to refresh every sub-project under an umbrella (not just the missing ones), pass --all too:
node "${CLAUDE_PLUGIN_ROOT}/dist/bin/sb-discover.js" --force --all "<project_dir>"
What the binary actually does given <project_dir>:
- Blocked path (HOME, Documents, Library, /tmp, etc.) → silent skip, no note. Tell the user the path is on the blocklist and discovery cannot run there.
- No strong project signal (just a README or LICENSE, no real manifest) → silent skip. Tell the user the directory doesn't look like a code project.
- Single project → one note at
~/.superbrain/vault/projects/<slug>.md. If the note already exists, a fresh## SuperBrain discovery (date)section is appended. - Umbrella / monorepo (workspace declaration or ≥2 sibling sub-projects detected) → one note per detected child (capped at 8), each named
<umbrella-slug>-<child-slug>.mdwithumbrella: <umbrella-slug>frontmatter. Without--all, only children whose note doesn't exist yet are written; with--all, every child's note gets a fresh appended section. - Sub-project of an umbrella, opened directly → discovery treats it as a single project but with umbrella context, so the slug is correctly prefixed.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 2d ago First seen · 59 lines · 84 tokens per session scan A e8b0b0fd75e7
discover is a command published in the GitHub repository m3talux/superbrain (41 stars, last pushed 22d ago), licensed MIT. It adds 84 tokens to every session and 1,020 once invoked, about $0.0004 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other commands, from other repositories
om-incident-capture
Capture an incident from Slack channels, DMs, and threads into structured vault notes. Produces a complete incident work note with timeline, people, analysis, and brag doc entry.
om-capture-1on1
Take a meeting transcript, notes, or Gemini summary and create a structured vault note with key takeaways, quotes, action items, and DM context.
om-weekly
Weekly synthesis — cross-session review of vault activity, North Star alignment, patterns, uncaptured wins, and forward priorities.
om-dump
Freeform capture mode. Dump anything — conversations, decisions, incidents, wins, thoughts — and I'll route it all to the right notes with proper templates, frontmatter, and wikilinks.
okf
Command "okf" from kimsanguine/llm-brain, covering 인자 파싱, 기본 private export, 🔴 share-ready gate (외부 공유 전 one-way door), legacy step 1: private dry-run 검토 and legacy step 2: private export.
pdf-to-wiki
Drive에 있는 대용량 PDF를 NotebookLM에 외주화하여 분석하고, 결과를 OUTPUTDIR에 Obsidian 노트로 저장한다.