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/nicelight/devrails-26/mb-syncgit clone --depth 1 https://github.com/nicelight/DevRails-26What 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 | $0.00018 | $0.01189 |
| Opus 5 | $0.00009 | $0.00594 |
| Sonnet 5 | $0.00004 | $0.00238 |
| Haiku 4.5 | $0.00002 | $0.00119 |
Grade A, and why
mb-sync 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 — 115 lines — stays where its author put it; the contents beside it link to each section on GitHub.
/mb-sync - Reconcile durable Memory Bank state
<input_contract>
Read and follow .memory-bank/workflows/mb-sync.md plus only the changed or
referenced Memory Bank/task/evidence surfaces needed for this boundary.
Normal boundary is once at the end of the current wave, after the scheduler or
explicit owner has written every closure/failure/blocking decision, final task
status, and evidence link to indexed .task.json records. Early sync requires
a recorded current-wave dependency on reconciled RTM/index/spec/contract/
changelog state or an explicit owner request; it does not replace the final
wave-boundary sync.
In manual flow, the invocation or its durable context identifies the explicit top-level caller/owner that will own applicable post-sync gates and the next handoff.
Do not require full sync for a local manual T0/T1 closure when only task
status, task verify, and compact .protocols/<TASK>/run.md changed.
</input_contract>
<hard_invariants>
- JSON task records remain authoritative for task status, dependencies, tier, gates, verification targets, and evidence markers.
- Status/closure/promotion ownership comes from
.memory-bank/workflows/tier-policy.md#tier-obligations,.memory-bank/workflows/tier-policy.md#closure-authority, and.memory-bank/workflows/mb-sync.md. /mb-syncnever infers or writes closure/failure/blocking/promotion,planned -> ready, dependent unblock, or dependent block when the explicit owner decision is absent.- In scheduler mode, sync only state already written by
/autopilotor/autonomous. In manual mode, require an already-recorded explicit owner decision or a direct instruction for this sync. spec-index.mdremains a pure registry; no decision bodies, feature status maps, ownership, or reverse usage are added.- Do not create a new task/status/boundary/design lifecycle or guess stale product/design state. </hard_invariants>
<operator_decisions> If authoritative artifacts disagree, or reconciliation exposes a new product, design, canonical-path, contract/state/data/storage/security/compatibility, task-boundary/tier/dependency/verification, closure, or promotion decision, stop with a consistency gap.
Interactive flow asks the operator/explicit owner and requires the answer to be
written to the existing owning artifact before sync resumes. Unattended flow
returns the gap to the scheduler for HALT_CLARIFICATION_REQUIRED,
HALT_BLOCKING_QUESTIONS, or HALT_QUALITY_GATES as applicable. A
recommendation/default is not an owner decision.
</operator_decisions>
<agent_discretion> Within the canonical sync checklist, the agent chooses inspection order, tools, and the minimum changed surfaces needed to prove consistency. It may repair mechanical links/routers and reconcile already-decided state, but may not create missing product/design/lifecycle decisions. </agent_discretion>
<required_outputs> As applicable, reconcile:
- task index/records and evidence links;
- requirements RTM and REQ lifecycle;
- epic/feature lifecycle separately from document status;
- truthful existing spec-backbone, spec-index, feature design links/status, and changed canonical specs without inventing design;
- accepted
boundary-map.mdlinks after path/heading changes; - behavior-spec links as optional examples only;
- Memory Bank/root/subfolder routers;
.memory-bank/changelog.mdfor the current wave/change.
Report consistency gaps and promotion eligibility without applying scheduler-owned promotion or final lifecycle transitions. </required_outputs>
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 · 115 lines · 18 tokens per session scan A fabd431afa9d
mb-sync is a command published in the GitHub repository nicelight/DevRails-26 (22 stars, last pushed 5d ago), licensed MIT. It adds 18 tokens to every session and 1,189 once invoked, about $0.0001 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
git
Git operations with intelligent commit messages and workflow optimization.
checklist
Generate a custom checklist for the current feature based on user requirements.
clarify
Identify underspecified areas in the current feature spec by asking up to 5 highly targeted clarification questions and encoding answers back into the spec.
specify
Create or update the feature specification from a natural language feature description.
analyze
Perform a non-destructive cross-artifact consistency and quality analysis across spec.md, plan.md, and tasks.md after task generation.
converge
Assess the current codebase against the feature's spec, plan, and tasks, then append any remaining unbuilt work as new tasks to tasks.md so implement can complete it.