exe

A command for running one specifically identified JSON task within defined time, scope, and evidence rules. JSON is a structured text format commonly used to represent task data.

In plain words
What is it for?
Use it to execute an indexed engineering task after its dependencies, specifications, module boundaries, and required evidence have been resolved.
Why use it?
It prevents the command from choosing unrelated work and requires the task's records, constraints, and acceptance details before execution.

Command

Install

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.

agentmods
npx agentmods add commands/nicelight/devrails-26/exe
Clone the repo
git clone --depth 1 https://github.com/nicelight/DevRails-26
Per session 23 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 4,542 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What 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.

ModelPer sessionOnce invoked
Fable 5 $0.00023 $0.04542
Opus 5 $0.00012 $0.02271
Sonnet 5 $0.00005 $0.00908
Haiku 4.5 $0.00002 $0.00454

Measured 2d ago against content hash 32c8e8dd1fad, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

exe 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.

skills/_shared/references/commands/exe.md · 370 lines

How it starts

The opening of the file, as written. The whole thing — 370 lines — stays where its author put it; the contents beside it link to each section on GitHub.

/exe - Execute one TASK

<input_contract> Expected $ARGUMENTS: one concrete TASK-NNN-TN-FT-NNN-WN.

The caller has already selected this task. /exe never scans the queue or chooses another task.

Require and resolve:

  • .memory-bank/tasks/index.json and exactly one matching indexed task record;
  • .memory-bank/workflows/tier-policy.md#tier-obligations;
  • #task-claim-and-dependency-ownership;
  • #tier-classification-and-escalation only when observed scope may require a higher tier, #closure-authority only for manual T0/T1 closure, #hard-write-boundary only when a boundary/alias is non-empty, #task-scoped-acceptance-evidence when its fields are populated, and #claim-linked-red--green-for-t2t3 only for T2/T3;
  • the task's direct feature/REQ context needed to interpret its outcome;
  • direct task-linked canonical SDD specs and executable constraints;
  • current boundary-map.md module, dependency-graph, and contract blocks when the task changes or crosses a registered module boundary;
  • behavior specs only when linked in source_artifacts and useful as non-authoritative examples.

For a product task whose feature is not FT-000, also require a positive Global Backbone Planning Revision and the latest feature-specific /review-tasks-plan APPROVE report with exact standalone REVIEWED_PLANNING_REVISION: <N> equal to it. FT-000 keeps its dedicated Foundation gate and does not use product task-plan review.

Use task purpose, success_outcome, anti_goals, source_artifacts, normative_inputs, constraints, invariants, verification_targets, evidence_required, and runtime_context when present. For T0/T1, do not load broad global planning docs by default. For T2/T3, direct task-linked applicable canonical specs are mandatory; feature links and spec-index.md alone are not execution context.

Point-of-use preflight must confirm:

  • index/file/ID resolution and matching ID tier/feature/wave segments;
  • valid tier: T0|T1|T2|T3 and executable lifecycle state;
  • every dependency exists and is done|done_for_prod;
  • dependency outcomes are authoritative prerequisites; execute only the selected task's owned outcome and integration delta;
  • no recorded blocker or unresolved required gate in the resolved task context;
  • success is observable from AC/REQ/spec/gates/verification targets;
  • the selected task, its direct feature/REQ context, and direct task-linked canonical specs are mutually consistent;
  • for a product task, its feature lacks PLANNING_RECONCILIATION_REQUIRED and its reviewed and current Planning Revision are equal;
  • T2/T3 direct canonical coverage is applicable and concrete enough to avoid guessing shape, rules, errors, and verification;
  • every current accepted planned|ready T2/T3 task satisfies the prospective proof-path contract in tier policy;
  • actual work fits the semantic outcome/AC/REQ/spec boundary, tier, deliberate hard allowed/forbidden scopes, and stop conditions;
  • when linked rules define architecture boundaries, the tactic keeps state changes and cross-module orchestration with their accepted owners, uses required public boundaries, preserves source-of-truth and dependency direction, creates no unaccepted cross-module contract or forbidden command/write bypass, and honors linked orchestration-placement rules;
  • for a task that changes a provider, current graph traversal covers every affected consumer until a compatible contract stops propagation; uncovered consumers make the feature tasking surface stale;
  • within an open domain-modeling choice, prefer the affected area's local style; material boundary or ownership changes use the existing operator-decision routes;
  • before creating, moving, or renaming a project-authored source file, resolve its applicable ownership, reuse, or tooling boundary; inspect the nearest relevant local pattern and governing manifests, configuration, resolution, or registration; preserve framework placement, required or reserved filename parts, import/module identity, public path, package export, route, and build target; then choose the minimum complete path with sufficient durable context;
  • for an Alembic migration under an accepted linear topology, preflight may resolve the current head transiently to set or verify the new revision's direct down_revision; do not propagate it as a repository-current-head assertion or update historical feature tests, and run the existing project-level graph contract unchanged;
  • the tier-required existing protocol is coherent, or every missing protocol file can be initialized from its framework-owned template before task start.

Read the full file on GitHub · 370 lines

Changes

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.

  1. 2d ago First seen · 370 lines · 23 tokens per session scan A 32c8e8dd1fad

Subscribe to this mod's changes

exe is a command published in the GitHub repository nicelight/DevRails-26 (22 stars, last pushed 4d ago), licensed MIT. It adds 23 tokens to every session and 4,542 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.