agentdesk

A system for managing complex software projects through task cards, role-based work, shared records, and explicit acceptance steps. It supports several independent agents working across separate development workspaces.

In plain words
What is it for?
Use it to coordinate planning, consultation, implementation, review, handoffs, and integration for software projects. It can also organize architecture, requirements, risk, and product discussions.
Why use it?
It makes multi-session work recoverable and shows who did what, what was delivered, and whether the result was accepted. It also provides a structured way to gather different viewpoints before implementation.

Skill for Claude CodeCodex

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 skills/levixdd/agentdeskskill/agentdesk
Any agent
npx skills add LeviXDD/AgentDeskSkill --skill agentdesk
Clone the repo
git clone --depth 1 https://github.com/LeviXDD/AgentDeskSkill

Made for: Claude Code, Codex.

Per session 96 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 8,002 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 $0.00096 $0.08002
Opus 5 $0.00048 $0.04001
Sonnet 5 $0.00019 $0.01600
Haiku 4.5 $0.00010 $0.00800

Measured 2d ago against content hash 3349ba90d25c, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

agentdesk 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.

The scan reads SKILL.md. This mod also ships 8 executable files (scripts/adopt_legacy_project.py, scripts/check_compatibility.py, scripts/init_project.py, …), listed below but not scanned — reading those needs a real analyzer, not pattern matching.

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/agentdesk/SKILL.md · 402 lines

How it starts

The opening of the file, as written. The whole thing — 402 lines — stays where its author put it; the contents beside it link to each section on GitHub.

AgentDesk

Outcome

Turn a software repository into a controlled multi-session development system:

Versioned task cards, PM-owned canonical state, real independently visible role tasks, role-bound execution, proactive callbacks, immutable Git evidence, explicit acceptance, and recoverable handoff.

The observable loop is non-negotiable:

PM/Leader task
→ verified independent role task named `role_no . role_name`
→ real dispatch delivery
→ dedicated worktree execution and durable report
→ worker proactively messages the PM/Leader source task
→ PM independently accepts and integrates
→ dependencies unlock
→ heartbeat reconciles only if the primary callback path fails

Repository records make this loop recoverable; they do not substitute for creating, addressing, or messaging the real tasks.

When a question needs competing viewpoints before execution, AgentDesk can prepend a bounded consultation loop: one Leader opens one to three read-only consultation tasks, collects independent first-round callbacks, optionally runs one cross-check round, synthesizes disagreements, and asks the Owner to confirm what—if anything—becomes a decision or task card. Consultation is advisory and never substitutes for the execution loop above.

Automate coordination, not authority. Keep product decisions, risky approvals, merging, and release behind their declared gates.

Resolve The Requested Mode

Choose the primary mode from the request and repository state. Consultation may precede PM, but its output does not silently enter execution:

  1. Initialize — scaffold or upgrade a project workflow.
  2. Consultation — open zero to three bounded, read-only advisory tasks around one topic, collect their callbacks, synthesize, and request Owner confirmation.
  3. PM — clarify requirements, configure roles, create and dispatch cards, review deliveries, update state, and unlock dependencies.
  4. Worker — implement one assigned card inside its scope and produce a delivery report.
  5. QA / Integration — test a frozen candidate or combine accepted commits under a dedicated card.
  6. Validate — check project structure and state without changing business code.
  7. Recover — reconcile conflicting state, lost callbacks, stale revisions, expired leases, or invisible Git evidence.

Read the full file on GitHub · 402 lines

Files

What ships with it

46 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.

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 · 402 lines · 96 tokens per session scan A 3349ba90d25c

Subscribe to this mod's changes

agentdesk is a skill published in the GitHub repository LeviXDD/AgentDeskSkill (10 stars, last pushed 1mo ago), licensed MIT. It adds 96 tokens to every session and 8,002 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.