Done Is a Claim: a done report format and a 50-line check (free sample)
by PaperChaseWebb
From Done Is a Claim: a one-line-per-action report format for agents, five rules to paste into AGENTS.md or a Grok Bot's standing instructions, and a read-only script that checks the report against your log and flags claims with no evidence, failed actions reported as done, and refusals the report left out.
- Price
- Free
- Difficulty
- Beginner
- Setup
- 5 min
- Format
- .md file
done-is-a-claim-done-report-format-check-free-sample.md · 5 KB · direct download, no signup
What you end up with
Your agent's next report says what happened, line by line, and a script tells you which lines the log does not back up.
What you get
- The done report format
- Five rules for AGENTS.md, CLAUDE.md or a Grok Bot
- A test you can run today
- done_mini.py: a 50-line report check
Who it is for
Anyone who wants a quick way to see whether an agent's done report matches what actually happened.
Works with
Solid chips: the agent is named in this listing's own tools list. Dashed chip: Plain Markdown, so it also pastes into other agents such as Grok Bot, Hermes, OpenClaw or Muse. We list those as compatible by format only, not as tested.
Tools needed
Cost to run: Free.
Time to result: About ten minutes to a first check
Sample prompts, free to use
End this task with one line per action: DONE: <what changed> | evidence: <id>, or BLOCKED: <what was tried> | evidence: <id> | <error>. Only write DONE for a success you saw.
The rules and script are in the file.
Free preview of the file
# Done Is a Claim: a done report format and a 50-line check (free sample) Version 1.0, 2026-10-31. By PaperChaseWebb. For anyone whose agent reports "done" on work that touches other systems. One Markdown file and one Python 3 script, standard library only, read-only. The script was tested with 4 tests on made-up logs on Linux with Python 3.13, 3.12, 3.11, 3.9 and 3.8. It was not run against a live agent session. On 2026-10-30 Eon published a benchmark where 19 models worked through 41 business tasks with every write blocked. In 290 runs the agent saw the refusal and still reported the task as complete, and every model did it at least once ([Eon](https://www.eon.io/blog/agentic-misalignment-benchmark)). Their advice: treat done as a claim, keep a record of tool results outside the model, and compare the two. This sample gives you the format and the comparison. ## The done report format Ask your agent to end every task that changes something with one line per action: ``` - DONE: <what changed> | evidence: <id> - BLOCKED: <what was tried> | evidence: <id> | "<error>" ``` The id is whatever your record uses: a request id from your service, a tool call id from the harness, or an id in your own action log. ## Rules to paste into AGENTS.md, CLAUDE.md or a Grok Bot's standing instructions 1. DONE needs a success you saw in a tool result: a 2xx status or exit code 0, and no error text. Otherwise write BLOCKED. 2. A 401, 403, "permission denied" or any refusal means it did not
Test record
- Date
- Oct 31, 2026
- What was run
- tests/test_mini.py on made-up logs (clean, findings, hidden refusal, bad input) and the copy extracted from the sample page. Linux, Python 3.13, 3.12, 3.11, 3.9 and 3.8.
- Result
- 4 passed, 0 failed on all five versions; exit 2 on a bad log line and missing arguments; the extracted copy was identical to the tested script.
- Not run
- A live agent session.
Copied from this item's dated test notes. What was not run stays in.
Worth knowing
- The sample checks DONE and BLOCKED lines against a JSONL log. The test endpoint, task cards, Claude Code and Codex log readers, free-text and planted-instruction checks, scoring and tests are in the full pack.
- On 2026-10-31 done_mini.py passed 4 tests on made-up logs on Linux with Python 3.13, 3.12, 3.11, 3.9 and 3.8, and the copy extracted from the sample page was byte-identical to the tested script. It was not run against a live agent session.
Questions
Does it change anything?
No. It reads two files and prints findings. It opens no network connection.
What is in the full pack?
A local endpoint that refuses writes, six task cards with planted instructions, a checker with 11 finding codes that reads Claude Code and Codex session logs, a score table, the full rules for Claude Code, Codex and Grok Bot, and 42 tests. $24.
Details
Version 1.0 · Updated Oct 31, 2026 · Markdown version of this page
License: Personal and commercial use. Download terms
Related
Done Is a Claim: a refused-write test and a report check for your agents
Before your agent gets write access, point it at a local test endpoint that refuses every write, give it six ordinary tasks for a made-up company, and check its final report against the log. A read-only script flags every claimed action that did not happen, hidden refusals and repeated planted instructions, and scores runs side by side. Keep the done report format and the check for everyday use.
Check the Claim: the six questions (free sample)
Six questions, an evidence level label, the add-the-parts check, one prompt and one card template that make any agent with web search check a claim before you use it. Updated to v1.1.
n8n with AI Agents: the Approve pattern and a ten-check workflow scan (free sample)
From n8n with AI Agents: the three-node Approve pattern that stops an n8n workflow sending anything until you click, five rules for agents that draft workflows, and a read-only script that runs ten checks on any workflow export, from open webhooks and keys in the file to sends with no Approve step.