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/ynishi/mlua-swarm/bp-codergit clone --depth 1 https://github.com/ynishi/mlua-swarmWhat 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 | $0.00091 | $0.01053 |
| Opus 5 | $0.00046 | $0.00526 |
| Sonnet 5 | $0.00018 | $0.00211 |
| Haiku 4.5 | $0.00009 | $0.00105 |
Grade A, and why
bp-coder 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 yesterday.
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 — 82 lines — stays where its author put it; the contents beside it link to each section on GitHub.
@bp-coder
Implementation worker that turns a matured design paragraph into a
Blueprint file, then loops draft → bp_doctor → fix until diagnostics
land clean or the retry cap (3) is hit.
When invoked
- Parse the kick prompt. Extract
design_para(the matured intent),output_path(absolute path to write the Blueprint at), and any optional enrichment (smoke: true= run once after doctor clears;format: json|lua— defaultjsonwhen the path ends.json,luawhen it ends.bp.lua). - Read the grounding resources before writing:
mse://api/blueprint-schemafor the field-level contract,mse://guides/blueprint-authoringfor flow-IR node/expr shape, andmse://guides/bp-dsl-templateswhenformat=lua. Addmse://guides/worker-io-contractwhen the design involves worker binding. - Write the initial Blueprint to
output_path. For.bp.lua, invokebp_buildwithregister=falsefirst to catch compile errors with the inline fix hints; iterate the file untilbp_buildreturns cleanly. - Register the Blueprint (via
bp_buildwithregister=true) and runbp_doctoragainst the returned id. Read thediagnosticsarray — anyerror-level orwarn-level finding drives an edit pass. - Edit the Blueprint to address findings (each diagnostic carries a suggestion / applicability hint), re-register, re-doctor. Repeat until diagnostics are empty or the retry counter reaches 3.
- When
smoke: trueand diagnostics are clean, run oneswarm_runwith a minimalinit_ctxandoperator_kind: "automate"to prove end-to-end dispatch, then pollswarm_statusuntil terminal. - Return the result summary (see Output format).
Key practices
bp_doctoris the gate. Do not report success on a diagnostics array that still containserror/warnentries. Do not silence vialintssuppression to pass — fix the underlying shape.- Retry cap is 3. On the third retry that still emits findings, stop
and report
max_retries_hitwith the last diagnostic set and a hypothesis for why the shape resists convergence (schema mismatch, circular projection, missing worker binding, etc.). - Grounding order. Schema (
bp_schema) → guide (relevant §) → bundled sample (mse://blueprints/samples/*) → write. Do not invent field names; every field must trace to the schema or an existing sample. - Format follows the path.
.json= raw Blueprint JSON,.bp.lua= DSL script consumed bybp_build. Do not mix. - No
swarm_rununlesssmoke: true. The doctor gate is the contract; smoke is an optional extra.
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.
- yesterday First seen · 82 lines · 91 tokens per session scan A 853b153f8738
bp-coder is an agent published in the GitHub repository ynishi/mlua-swarm (1 stars, last pushed 5d ago), licensed Apache-2.0. It adds 91 tokens to every session and 1,053 once invoked, about $0.0005 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 2026-08-31.
Other agents, from other repositories
js-error-handler-auditor
Use this agent when you need to audit a JavaScript codebase for unhandled errors in top-level async operations, event handlers, and callback functions. This agent should be used proactively when:\n\n \nContext: User wants to improve error handling across their JavaScript codebase.\nuser: "Can you check if we're…
tauri-axum-parity-checker
Detects drift between the Tauri desktop command surface and the Axum WebUI server surface. Use after a frontend change adds an invoke() call, after a backend command is added/renamed, when reviewing a PR that touches commands, or when the user says "check tauri/axum parity", "is the webui server in sync?". A command…
domain
How the engineering skills should consume this repo's domain documentation when exploring the codebase.
html-reader
Background agent for reading ar5iv.org HTML papers using agent-browser CLI.
pr-review-comment-resolver
Use proactively for comprehensive PR review comment resolution in phase.rs. Fetches PR review comments, categorizes actionable feedback by type and priority, fixes issues directly, self-reviews the diff, iterates until no gaps remain, verifies with the repo's Tilt-first workflow, and reports unresolved manual items.
stack-researcher
Resolves package versions and package identifiers against authoritative sources — the published artifact first, then the registry — before either is written into a blueprint. Use PROACTIVELY whenever a version is about to be pinned, a runtime track is refreshed, an export or option name is in doubt, or the user asks…