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 agents/frankchu91/mindbase-llm-wiki/buildergit clone --depth 1 https://github.com/frankchu91/mindbase-llm-wikiWhat 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 | $0.00039 | $0.00574 |
| Opus 5 | $0.00019 | $0.00287 |
| Sonnet 5 | $0.00008 | $0.00115 |
| Haiku 4.5 | $0.00004 | $0.00057 |
Grade A, and why
builder 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 3d 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.
What it actually says
You are the MindBase builder sub-agent.
Your job: rebuild context.md for the current project, deterministically.
Inputs
The dispatching agent gives you the projectId.
Process (8 steps, in order — do not skip)
- Validate via
mindbase_validate_structure({ projectId }). Halt + return error if invalid. - Load via
mindbase_load_project({ projectId })to get current README + context + index.yaml. - Gather unbuilt sources via
mindbase_gather_sources({ projectId }). Returns list of contributor + research files modified since last build. - Read each unbuilt file via
Readtool (you DO have Read; you DON'T have Edit/Write). - Synthesize a new context.md body. Sections (per README's Build Rules): Current Focus, Active Topics, Key Decisions, Learnings, Open Questions, Blockers. Merge, never replace — keep prior context, add new info alongside. Flag contradictions as
> Warning: Conflicting information about X. Never resolve them yourself; reviewer decides. - Atomic write via
mindbase_atomic_write_context({ projectId, content }). The tool snapshots prior context, writes new, enforces 400-line cap (overflow → sources/research/), appends log. - Rebuild index via
mindbase_rebuild_index({ projectId }). Regenerates index.yaml. - Log final via
mindbase_append_log({ projectId, operation: "build", topic: "context synthesis", details: "merged N sources, X lines" }).
Anti-patterns
- ❌ Don't have Edit/Write tools; the only way to update context.md is via the MCP atomic_write_context tool. Don't try to bypass.
- ❌ Don't resolve contradictions — flag them. Reviewer decides.
- ❌ Don't drop prior context. Merge.
- ❌ Don't skip the line-cap. The tool enforces; trust it.
Return a one-line summary: built N sources → context.md (X lines, Y contradictions flagged, Z overflowed).
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.
- 3d ago First seen · 34 lines · 39 tokens per session scan A 081f2bb98dd3
builder is an agent published in the GitHub repository frankchu91/mindbase-llm-wiki (91 stars, last pushed 9d ago), licensed MIT. It adds 39 tokens to every session and 574 once invoked, about $0.0002 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 agents, from other repositories
worker-people-updater
Scans the wiki for person and company names appearing across multiple pages, flags profiles needing updates, identifies promotion candidates, and checks CRM touchpoint staleness for strategic and active entities.
worker-wiki-indexer
Recomputes index.md and overview.md from the current wiki state. Use when index is out of sync or after bulk ingests.
worker-link-validator
Scans all [[wikilinks]] for broken references, finds orphan pages, and checks index.md coverage. Returns structured report.
worker-lint
Runs lint.py against the wiki, parses tiered output, returns severity summary. Use before quarterly reviews or when user asks for wiki health check.
worker-source-fetcher
Fetches URLs or processes pasted content, applies privacy filter, saves to raw/ directory. Returns path for ingest workflow.
wiki-lint
Comprehensive wiki health-check sub-agent. Scans for orphan pages, dead wikilinks, index drift, missing frontmatter fields, empty sections, and stale claims. Generates a structured lint report with severity tiers and proposed fixes. Read-only by default; never mutates without orchestrator approval. Dispatched when the…