Template Safety Card: what it can touch, what needs your yes, what it was tested on
by PaperChaseWebb
One honest page for every agent template you sell or share: what it can touch, what waits for the buyer's yes, what it never does, which secrets it needs and what it was really tested on. A checker fails the template when the card is missing or does not match the files: a network call the card never mentions, texts sent with no yes, a key name left off, 25 prompts claimed for 3, a test row with no saved log, a card from an older version. It checks your listing against the card too. You approve each card at your own terminal.
- Price
- $24
- Difficulty
- Intermediate
- Setup
- 30 min
- Format
- .md file
template-safety-card-what-it-touches-what-needs-your-yes-tested-on.md · 216 KB · instant download after checkout
Want this built for you instead? We can build it for you.
What you end up with
Every template you release carries a card that matches its files, its counts and its test logs, your listing says nothing the card cannot back up, and a change to any file shows up as a stale card before a buyer sees it.
What you get
- 21 sections and an appendix in one .md file, with a one-line unpack for all 103 files
- The seven-part card: What it can touch, What needs your yes, What it never does, Secrets it needs, Tested on, Not tested and What is inside, tied to the template's file hash
- card_check.py scan and draft: what the files can touch, every host and secret name with file and line, and a first card to finish
- card_check.py check: 15 card checks, 10 hazard checks (keys in files, curl into a shell, flags that turn off asking, hidden instructions) and 13 checks of the card against the files, from a touch left off to a padded count and a test row with no log
- card_check.py listing: 10 checks on your sales copy, such as agents nobody tested, "works offline" over a network call, "no API key needed", "fully autonomous", safety promises and demand claims
- card_check.py fix: 10 checks so an agent can only add and correct, never take out a yes row, add a test row without a log or raise a count
- Report and rules checks, and approve.py: it needs a real terminal and part of the card's hash, refuses agent sessions and piped input, and checks the card has not changed before it goes into the template
- An eval set with 10 made-up templates, 3 listings and 2 reports, rules and two skills for Grok Bot, Claude Code, Codex and Hermes, six prompts and templates
- The test harness: 109 unit tests and 280 planted faults you can repeat
Who it is for
Template sellers and agencies who want one checked page that tells buyers what a template can touch, what needs a yes and what it was tested on, with a person approving every card.
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: The kit is the only purchase. Everything runs locally with Python; the checker reads files only, runs none of them, makes no network calls and logs in to nothing.
Time to result: The worked example in 5 minutes; a first card for your own template in about 30 minutes
How it works
Unpack and test
Save the kit as pack.md in an empty folder, run the unpack line from Appendix A, then sh test_all.sh.
Expected: 103 files written; 109 tests OK; 280 of 280 planted faults caught; the eval set and rules check pass.
Watch the flow
Run sh example/run_example.sh and read it next to section 17.
Expected: A made-up template's card with 12 findings, the agent's first fix caught with 3, the second fix clean, a listing with 9 findings, a first-try report with 3 and a piped approval refused.
Card your template
Copy your template into ./template, edit settings.json, then run python3 desk/card_check.py scan template and python3 desk/card_check.py draft template.
Expected: A list of what your files touch and a first card to finish.
Check, fix and approve
Have your agent follow skills/safety-card-write and skills/listing-check, run card_check.py fix until 0 high, then approve the card at your terminal and upload the template yourself.
Expected: A card with 0 high findings in your template and a listing that matches it.
Sample prompts, free to use
Read AGENTS.md. Run card_check.py scan template and check template, and tell me in plain words what this template can touch and what the card gets wrong. Change nothing.
Prompt 1 of 6.
There is no card yet. Run card_check.py draft template, then follow skills/safety-card-write/SKILL.md and write queue/fixes/SAFETY-CARD.md. Leave every test row I have not given you a log for under Not tested.
Prompt 2 of 6.
Follow skills/listing-check/SKILL.md for listing/[name].md and write the fixed copy to queue/fixes. List each change and the code behind it.
Prompt 4 of 6.
Free preview of the file
# Template Safety Card: what it can touch, what needs your yes, what it was tested on Version 1.0, 2026-11-01. By PaperChaseWebb. For people and agencies who sell or share agent templates (prompt packs, skills, scripts, agent files and standing orders for Grok Bot, Claude Code, Codex, Hermes and the like), who want one honest page in every template that says what it can touch, what waits for the buyer's yes, what it never does, which secrets it needs and what it was really tested on, and a checker that fails the template when that page is missing or does not match the files. Works with Grok Bot, Claude Code, Codex and Hermes. Python 3.8 or newer, standard library only, no API keys, no logins. The checker reads files; it never runs them, opens no network connection and changes nothing. Tested with 109 unit tests and 280 planted faults on Linux with Python 3.13, 3.12, 3.11, 3.9 and 3.8. All sellers, templates, contacts and test logs in this kit are made up. Buyers have started to ask before they install. A template that sends texts, reads a mailbox or holds a payment key can do real harm in the wrong setup, and builders on X now warn people not to install templates on sight. Third parties rescore templates from the outside. The seller, who knows the template best, usually writes nothing. When sellers do write something, it drifts. The listing says 25 prompts and the folder has 3. It says "works with Grok Bot, Claude Code, Codex and Hermes" and it was run once, in one of them
Test record
- Date
- Nov 1, 2026
- What was run
- tests/test_lib.py (7), test_card.py (17), test_match.py (24), test_listing.py (11), test_fix.py (11), test_claims.py (8), test_modes.py (9), test_set.py (4), test_mini.py (6) and test_approve.py (12, on a real pseudo terminal); tests/mutate.py with 263 planted faults across cards, template hazards, card against template, listings, bad fixes, reports and agent rules, plus 17 against card_mini.py; card_check.py set on the eval set; example/run_example.sh end to end; and the unpack line with test_all.sh. Linux, Python 3.13, 3.12, 3.11, 3.9 and 3.8.
- Result
- 109 passed, 0 failed on all five versions; 280 of 280 planted faults caught and all clean baselines clean; the eval set matched 15 of 15; the example caught the bad card with 12 findings, the first-try fix with 3 and the first-try report with 3, and refused piped approval; the unpack line wrote every file and test_all.sh passed there.
- Not run
- A live Grok Bot, Claude Code, Codex or Hermes session. Live marketplaces. Binary files, minified bundles and code fetched at run time. macOS or Windows. A security audit.
Copied from this item's dated test notes. What was not run stays in.
Worth knowing
- The agent rules were checked by the checker but not loaded inside a live Grok Bot, Claude Code, Codex or Hermes session for this release. The eval fixes were written by hand to the rules. Run card_check.py set evals with your own agent's fixes before you rely on it.
- Not a security audit. A card that passes says the card matches the files; it does not say the template is safe. The scan reads text files and known patterns in Python, JavaScript, TypeScript, shell, JSON, YAML, TOML and Markdown; binary files, minified bundles and code fetched at run time are not read.
- The kit reads the listing you save as a file. It never logs in to a marketplace and cannot list, upload or update anything.
- The owner step assumes the agent cannot run approve.py at a terminal. Run it as a separate user; agent rules alone are not a hard boundary.
- All sellers, templates, contacts and test logs in the kit are made up. Tested on Linux only; macOS and Windows were not tested.
- On 2026-11-01 on Linux: 109 unit tests and 280 of 280 planted faults passed on Python 3.13, 3.12, 3.11, 3.9 and 3.8; the eval set matched 15 of 15; the unpack line wrote every file and test_all.sh passed in that folder.
Questions
How is this different from Vet Before You Install?
Vet Before You Install is for the buyer checking someone else's tool. This kit is for the seller: you write one card for your own template, and the checker makes sure it matches your files, your counts, your test logs and your listing.
Which agents does it work with?
Any agent that can read files and run a script. The kit has rules for AGENTS.md and CLAUDE.md, Claude Code and Codex notes, Hermes notes and a Grok Bot standing order. Those were checked by the checker, not loaded in each agent for this release.
Does it run my template's code?
No. It reads the files as text. It runs none of them, opens no network connection and changes nothing outside the queue folder.
Can my agent mark the template as tested?
Only with a saved log in the template. A Tested on row without a log fails, and the fix check refuses a changed test result without a new run.
Can you set it up for us?
Yes. Our Install sprint writes and checks a card for each of your templates, saves your test runs as logs and rewrites your listings from the cards, or Managed Ops checks every release for you. Book a call or write to [email protected].
Details
Version 1.0 · Updated Nov 1, 2026 · Markdown version of this page
License: Use inside your own team or company, including for clients and for templates you sell. Do not resell or republish the kit itself. Download terms
Related
Template Safety Card: the seven-part card and a ten-check card scan (free sample)
From Template Safety Card: the blank seven-part card, ten rules to paste into your agent, and a read-only script that runs ten checks on one template folder, from no card and a missing section to a network call, a send or a payment the card never mentions, a flag that turns off asking, a key in the files, a test row with no date or numbers and a safety promise.
Vet Before You Install: twelve checks, a sandbox recipe and a tested vetting folder for third-party agent code
Decide whether a skill, plugin, mod, MCP server, connector, repo or model gets into your agent setup: sort the seat, read every file that runs, map reach and the data path, pin and vendor it, trial it in a sandbox with a planted fake key, then record a decision the owner approves at their own terminal. v1.1 adds a seat sort before any vetting, two new checks (data path and reach, approval surfaces), evidence tiers, twelve token classes including hidden text, encoded strings, startup file edits and approval hooks, a fresh-session second read, data path questions for proxies and routers, dependency and model weight pinning, a fifth decision (not vettable) with recheck dates and an approved shelf, more special cases, a vetting folder with a fifteen-check read-only checker and an owner approval helper, agent files for Grok Bot, Claude Code, Codex and Hermes, and twelve new prompts (27 in all). Price unchanged.
Ask-first rules for AI agents: hard nos, approval cards and a test set
Paste-in rules that make an agent ask before it sends, spends, posts or deletes, define what a real yes is, and stop retries after a no, plus 12 test scenarios to check it.
Agent Skills Workshop: write, test and run SKILL.md skills for Hermes, Claude Code and Codex
A bot template and teaching guide for building your own agent skills instead of collecting other people's. A skill card, a three-skill split for long jobs, a corrections-to-skills loop, an intake check for skills you did not write, four action classes, a SKILLS.md register, an overnight run sheet, eight skill drills, a rules block and 14 prompts. Includes skill_check.py, AGENTS.md with six owner stops, two skills and templates.
MCP Server Kit: a tested starter server, approval-gated write tools and a checker
A small working MCP server for a made-up plumbing business where every write is a draft plus an apply that waits for the owner, an approval script the agent cannot run for you, and a read-only checker that finds unsafe tools, poisoned descriptions, open schemas, live keys and unpinned packages in any server or agent config, including starters you bought. Tested against the official MCP Python SDK with configs for Grok Bot, Claude Code, Codex and Hermes.
Check the Claim (expanded edition): a method, prompts and a skill for checking AI claims
Make your agent check income, tool, popularity, quote, news and offer claims before they reach a brief, a pitch, a purchase or a post: label how much was read, run the creator's own arithmetic, trace sources back to first hand, and give each claim one of five verdicts on a card the owner approves. v1.1 adds an evidence level label (full, caption, summary or wall), four new arithmetic tests, a section on tracing a source back to first hand, screens for offers, quotes and slogans, four more red flags, a sixth worked example, a claim folder with a fourteen-check read-only checker and an owner approval helper, agent files for Grok Bot, Claude Code, Codex and Hermes, and twelve new prompts (26 in all). Price unchanged.
AGENTS.md, CLAUDE.md and Cursor rules starter files
Six fill-in instruction files for coding agents (base, Next.js plus Supabase, FastAPI, browser game, CLAUDE.md pointer, scoped Cursor rules), an Approvals section that makes the agent stop and wait for your yes, a table of what Grok Bot, Claude Code, Codex, Hermes and Cursor each read, a read-only checker for your rule files and 9 prompts to write, trim and test your own.