Desk Authority Map: who in your agent fleet may decide what
by PaperChaseWebb
Four plain files say, for every desk and action, decide, propose, ask-owner or never, with dollar limits, escalation and the evidence each desk owes. Agents ask authority.py before they act and get the answer as an exit code; anything not in the map is never. Owner yeses are typed at the owner's terminal, tied to one request's sha256, good for one use, and refused inside an agent. Fifteen read-only checks run on the map itself.
- Price
- $29
- Difficulty
- Intermediate
- Setup
- 30 min
- Format
- .md file
desk-authority-map-who-may-decide-what.md · 75 KB · instant download after checkout
Want this built for you instead? We can build it for you.
What you end up with
A written answer to who may send, spend, delete, publish or change access in your fleet, which your agents check before acting, and which turns every answer to never if anyone edits it without your yes.
What you get
- 19 sections in one .md file
- The map folder: fleet.json, desks.csv, actions.csv and authority.csv, with blank templates and fourteen starter actions
- Four levels, three risk classes, dollar limits, escalation and the evidence rule
- authority.py: check (exit codes 0, 10, 11, 12), request, owner approval tied to a sha256, verify, and one use per yes
- authority_check.py: fifteen read-only checks, with a seeded test map (15 findings, then 0)
- A skill file for Claude Code, Codex and Hermes, an AGENTS.md snippet and a Grok Bot rules block
- Ten prompts, a failure table and a worked example
- A look at eleven governance packs and templates, with sources
Who it is for
Anyone running several agents who wants it written down, and checked, which agent may do what alone, what needs a second look, and what only the owner decides.
Works with
Solid chips: the agent is named in this listing's own tools list. Dashed chip: A skill file for Claude Code, Codex and Hermes, an AGENTS.md snippet and a Grok Bot rules block; the scripts were tested on Linux with Python 3.13, the agent files were not loaded inside each harness.
Tools needed
Cost to run: The pack is the only purchase. The scripts use no model tokens.
Time to result: A ratified map for a fleet of four to eight desks in an afternoon
How it works
Run the tests
Copy the desk-authority-map folder and run sh test_all.sh.
Expected: 15 findings on the seeded map and 0 on the fixed one, 15 of 15 single faults caught alone, checker tests 21 passed, authority tests 39 passed, exit 2 on a missing folder.
Draft your map
Run authority.py init, fill the four files with prompts 1 to 3, starting every rule at never. Run authority_check.py until only A15 remains.
Expected: Every desk has a human owner and a checker, money has limits, owner-only actions are ask-owner or never.
Ratify and connect
The owner runs authority.py approve DIR --map at their own terminal. Install the skill file, paste the AGENTS.md snippet, or paste the rules block into Grok Bot.
Expected: Each desk asks authority.py check before acting and follows the exit code.
Live with it
Ask-owner actions become requests the owner approves by hash. Review requests and never hits every quarter with prompt 9.
Expected: verify passes only for an unchanged request under the current map, within approval_days, once.
Sample prompts, free to use
List every action you have taken or could take with the tools you have right now. For each, say its kind, whether it can be undone and what it reaches. Write it as rows for actions.csv. Do not write any rules. Using desks.csv and actions.csv, draft authority.csv for the <desk> desk. Start every action at never. Loosen only where you can name the evidence you would leave.
Prompts 1 and 2 of 10.
Before <task>, list every action it needs and run authority.py check for each. Show me the answers in a table, then do only the decide ones.
Prompt 5 of 10.
Free preview of the file
# Desk Authority Map: who in your agent fleet may decide what Version 1.0, 2026-10-28. By PaperChaseWebb. A map of your agent desks written as four plain files, a script your agents ask before they act, a read-only checker with fifteen checks, templates, a seeded test map, ten prompts and the tests, all Python 3 standard library. Works with Claude Code, Codex and Hermes through a skill file and an AGENTS.md snippet, and with Grok Bot through a rules block. Tested as described in section 11: everything ran on Linux with Python 3.13, including owner approvals typed at a pseudo-terminal. The skill file, snippet and rules block were not loaded inside the harnesses for this release, and the pack has not been used on a real client fleet. Each is marked where it matters. Most agent teams have an org chart by now. A research agent, a support agent, a builder, a bookkeeper, each with a title, a tool list and a line to someone. The chart answers who does the work. It does not answer the questions that matter on a bad day. May the support agent send that email or only draft it? Can the builder buy a $300 tool on its own? Who checks the bookkeeper? What happens when the answer is "nobody wrote it down"? Spreadsheets and policy documents for this exist, and some are good. The weak spot is that nothing reads them at the moment an agent is about to act. This pack makes the map something an agent asks. `authority.py check` returns decide, propose, ask-owner or never as an exit code, and a
Test record
- Date
- Oct 28, 2026
- What was run
- authority_check.py on a seeded map and a fixed copy, each planted fault alone, a missing input, and the unit tests. Linux, Python 3.13.
- Result
- 15 findings on the seeded map, 0 on the fixed one; 15 of 15 planted faults caught alone; exit 2 on missing input; nothing written. 21 checker and 39 authority.py tests passed, 0 failed.
- Not run
- The agent files inside Claude Code, Codex, Hermes or Grok Bot. A real client fleet. Windows.
Copied from this item's dated test notes. What was not run stays in.
Worth knowing
- authority.py answers questions; it cannot stop an agent that never asks. The rules files tell each desk to ask, and the quarterly review catches drift, but this is not a sandbox.
- The skill file, AGENTS.md snippet and Grok Bot rules block were not loaded inside Claude Code, Codex, Hermes or Grok Bot for this release, so how reliably each harness asks first was not measured. The pack has not been used on a real client fleet. The test map is made up. Linux only; Windows was not tested.
- Not a staff AI usage policy and not a compliance certification (SOC 2, ISO 42001, EU AI Act).
- The approval step refuses piped input and the environment variables Claude Code, Codex, Hermes and Grok Bot set. That stops an agent approving by mistake; it is not a lock against someone with the same account.
- On 2026-10-28, on Linux with Python 3.13.5: authority_check.py gave 15 findings on the seeded map (each check once) and 0 on the fixed one, caught 15 of 15 single faults with only its own check, wrote nothing, and exited 2 on a missing folder, a missing fleet.json, desks.csv, actions.csv or authority.csv, or broken fleet.json; checker tests 21 passed, 0 failed; authority.py tests at a pseudo-terminal 39 passed, 0 failed.
Questions
How is this different from Governed Agent Teams?
Governed Agent Teams helps you design the team and includes an authority table in prose. This pack turns that table into rules an agent can ask at run time, with limits, escalation, a check on the map itself and owner approvals tied to hashes.
What happens with an action nobody mapped?
The answer is never, exit 12. New tools and workarounds start closed and get opened on purpose.
Can an agent approve its own request?
Not through the script. Approve refuses to run inside Claude Code, Codex, Hermes or Grok Bot environments or on piped input, and the yes is tied to the request's sha256, so an edited request needs a new yes.
What if I change the map?
Every answer becomes never until you ratify the new version, and requests written under the old map stop verifying.
Details
Version 1.0 · Updated Oct 28, 2026 · Markdown version of this page
License: Use on your own fleet and on fleets you set up for clients and teams. Do not resell or republish the pack itself. Download terms
Related
Desk Authority Map: five rules and a five-check script (free sample)
From the Desk Authority Map: five rules for an agent fleet (every desk has a human owner, no desk checks its own work, nothing that cannot be undone is decided alone, money gets a dollar limit, some actions belong to the owner), and a read-only script that checks one map file for them.
Governed Agent Teams: subagent personas, authority maps and handoffs
A Markdown pack for running a team of coding and research agents with clear limits: 7 subagent personas for Claude Code with Codex AGENTS.md and Hermes skill versions, a Desk Authority Map, three role risk classes, a handoff packet the receiver verifies, an evidence standard, a decision ledger and a reach inventory.
Agent Control Room: voice control for a dozen agents
Talk to twelve agents by name from one place. Say "tell juniper to fix the login test" and it goes to juniper and only juniper, read back in a sentence. Ask "who is waiting" and the board answers with waiting on top. Say "stop all" and every session gets it. What voice never does is approve: every yes, "approve it", "go ahead" and "1" is refused, and gated actions wait for a typed yes from the owner tied to the sha256 of the request, which the agent checks before it acts.
Agency Install Kit: install one agent workflow in two weeks, with a binder a script can check
Five paid packs and the free Readiness Self-Check, tied into one two-week install by a new Install Binder: plain files for readiness, one workflow, agent owners and tested pauses, a gate register, a reach inventory, a cost cap, a dry run with refusal cases, a scorecard with kill rules and a run sheet with an off switch. A fifteen-check script names every missing decision, and the owner's go-live yes is tied to the binder's sha256.
Production Readiness Audit: decide from evidence whether an agent system can go live
Twelve areas, each pass, gap or na, and every pass points to a dated evidence file. Gaps get a named owner and a due date, five areas block go-live until fixed, drills are run rather than described, and the verdict is signed by someone who did not build the system. A fifteen-check script reads the folder, the report is sealed to a hash of the evidence, and the owner's go-live yes is tied to the report's sha256.
Agent Readiness Self-Check: score one workflow before you hand it to an AI agent
A free self-check for agencies and small businesses. Name the decision, then score one workflow across eight areas (workflow mapped, data access, approval gates, owner per agent, handoffs, cost tracking, security and vetting, measurement) with an evidence rule, three gate rules and a skeptical pass, and read your band and the next steps it allows. v1.1 starts with the decision the score will change, scores unknowns as unknown, adds rules for conflicting facts and lookalike public records, ties every band to the next steps it allows with the Production Readiness Audit before go-live, makes owner approval concrete (prepared and approved status, a one-reply approval line, decisions the agent must leave open, a done test), adds agent files for Grok Bot, Claude Code, Codex and Hermes, a second worked example (missed call to booked job), a readiness folder with a fifteen-check read-only checker, a front door test with four questions to ask anyone selling you an agent install, and a plain note on how it was tested. Still free.
Rubric Library: name, version and own the scoring rubrics hiding in workflows
Every score, rank, flag or triage step an agent runs has a rubric in it, usually unnamed and rewritten each time. Six plain files give each rubric an owner, a version, grades with definitions and actions, evidence rules, borderline examples and an adversarial set. Agents load a rubric by name and number with rubric.py and check every score; unnamed or default is unverdicted, and anything touching a person, a client or money goes to a person. Fifteen read-only checks run on the library itself.
Executive Morning Brief: a daily brief for owners where every line has a source
A morning brief your agent builds from your own calendar, inbox, payments and tasks, read-only, with every line citing a record you can open and every figure copied from it. Sources that failed are named instead of hidden, a quiet day says so, and replies or calendar changes stay drafts until you say yes at your own terminal, tied to the sha256 of that day's brief. Payments and deletions are never drafted. Fifteen read-only checks run on the brief before it renders.
Reach Inventory: every connector your agents hold, with a job, an owner and a way out
One row for every connector, MCP server, OAuth grant, API key and automation bridge your agents can use, each tied to a job and marked necessary, convenient or leftover. Scope granted is set against the scope the job needs, keys are named by where they live and never pasted, bridges list every app behind them, and leftover access is removed with a date and proof. Agents list and propose; removals, narrower scopes, rotations and new grants wait for your yes at your own terminal, tied to the sha256 of the inventory. Fifteen read-only checks run on the inventory.