task-queue-mcp AGENTS.md

A local task server that lets software agents submit, find, retrieve, and update work items stored as YAML files.

In plain words
What is it for?
Use it to manage cross-agent tasks through a file-backed queue, including approval, progress updates, completion, failure handling, and cancellation.
Why use it?
It provides a shared queue so agents can dispatch work and track states such as approved, in progress, completed, failed, parked, or cancelled.

Instructions file for CodexOpenCode

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 instructions/tadmstr/task-queue-mcp/agents-md
Clone the repo
git clone --depth 1 https://github.com/TadMSTR/task-queue-mcp

Made for: Codex, OpenCode.

Per session 1,901 This file is loaded in full into every session.
When invoked 1,901 The same file — it is already loaded in full.
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 $0.01901 $0.01901
Opus 5 $0.00950 $0.00950
Sonnet 5 $0.00380 $0.00380
Haiku 4.5 $0.00190 $0.00190

Measured yesterday against content hash 2ea7ba91cd19, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

task-queue-mcp AGENTS.md 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.

AGENTS.md · 84 lines

How it starts

The opening of the file, as written. The whole thing — 84 lines — stays where its author put it; the contents beside it link to each section on GitHub.

task-queue-mcp

FastMCP server backed by YAML files on disk. Cross-agent task dispatch for the forge platform. Runs in a Docker container with the queue directory mounted.

What it does

Provides 10 tools for submitting, listing, retrieving, and mutating task files. Each task is a single YAML file (<timestamp>-<slug>.yml) in TASK_QUEUE_DIR.

Task status lifecycle:

submitted ──▶ approved ──────────▶ in-progress ──▶ completed
     │                                              └─▶ failed
     └──▶ pending-approval ──▶ approved            cancelled (terminal, any non-terminal state)
                                                    parked         (non-terminal, reversible)
                                                    routing-failed (non-terminal, dispatcher-written;
                                                                    operator can cancel/park/re-submit it,
                                                                    agents cannot reach it via update_task)

submitted ══▶ completed   (task_type=notify only — a creation path, not a transition;
                            written directly by submit_task, never via update_task.
                            VALID_TRANSITIONS["completed"] is still exactly {"in-progress"}.)

Terminal statuses: completed, failed, cancelled. Transitions are validated server-side; agents cannot claim (in-progress) a task that is not approved. The one exception is by construction, not transition: a notify task is written straight to completed at submit time and never passes through approved or in-progress at all.

Tools

  • submit_task(source_agent, target_agent, task_type, summary, description, ...) — Write a YAML task file to TASK_QUEUE_DIR (initial status submitted, except task_type="notify", which is written straight to completed — see the lifecycle diagram above). Returns {ok, task_id, filename}.
  • list_tasks(target_agent, source_agent, status, limit) — Filter tasks from the queue directory. include_archived and include_dead_letters are separate opt-ins, both off by default.
  • get_task(task_id) — Retrieve a single task by ID. Searches the queue root, then archive/, then dead-letters/.
  • update_task(task_id, status, ...) — Move a task to in-progress, completed, or failed (records actor/note/history).
  • set_task_status(task_id, ...) — Operator-lane transitions (approved, cancelled, parked, audited overrides, out-of-vocabulary repair). routing-failed is a valid non-terminal source here, same as any other — an operator can cancel, park, or re-submit a task stuck in it.
  • cancel_task(task_id, actor, note) — Cancel a non-terminal task.
  • park_task(task_id, actor, note) — Pause a task in place. Status-only; the file never moves.
  • unpark_task(task_id, actor, note, status) — Return a parked task to parked_from, or to an explicit status.
  • amend_task(task_id, amendment, actor, reason) — Append a correction under payload.amendments. Never mutates payload.description.
  • requeue_dead_letter(task_id, actor, note) — Operator-only. Move a record out of dead-letters/ back to the queue root at submitted. The only path in this server that walks a record out of a terminal status, and it is scoped to that one directory.

Read the full file on GitHub · 84 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. yesterday First seen · 84 lines · 1,901 tokens per session scan A 2ea7ba91cd19

Subscribe to this mod's changes

task-queue-mcp AGENTS.md is an instructions file published in the GitHub repository TadMSTR/task-queue-mcp (3 stars, last pushed 3d ago), licensed MIT. It adds 1,901 tokens to every session, about $0.0095 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.

Related

Other instructions, from other repositories