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.
npx agentmods add instructions/azrtydxb/procoder/testsgit clone --depth 1 https://github.com/azrtydxb/procoderWhat 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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00738 | $0.00738 |
| Opus 5 | $0.00369 | $0.00369 |
| Sonnet 5 | $0.00148 | $0.00148 |
| Haiku 4.5 | $0.00074 | $0.00074 |
Grade A, and why
procoder tests.instructions.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.
How it starts
The opening of the file, as written. The whole thing — 75 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Tests in procoder
A test here is not finished when it passes. It is finished when it has been watched to fail.
Every test carries its mutation
Each test has a // proved by: line naming a specific change to the
source that must make it fail:
// proved by: made the unmarshal failure return nil — a crashed semgrep
// then reads as a clean SAST pass.
func TestSastUnreadableOutputIsNotRun(t *testing.T) {
Then actually apply that change, build, and run the test. Three discoveries came only from doing it, never from reading the assertion:
- A test exercised a helper directly and never the dispatch, so deleting
the call from
run()left it passing. It goes through the real entry point now. - A P-CONTROL test ran
procoder formatonly over files that were already formatted, so the branch that prints a rewritten file never executed — and a mutation making that branch write to disk passed. A mutation must reach the branch; check the fixture can provoke it. - A mutation that does not compile is not a mutation, and neither is one that produces no diff. Verify it applied.
Where two redundant mechanisms guard one behaviour, no single mutation will fail the test. Say so in the comment and name the pair, rather than claiming a proof that does not hold.
Assertions
- Assert the verdict, not the subject.
grep "unformatted <file>", notgrep "<file>"— the file's name also appears in "UNCHECKED — csharpier is not installed", which reports the opposite of a finding. - Never assert on a word that appears either way. Checking for "deny" or "block" in a hook envelope passes on an allow, because "block" sits inside "1 blocking finding(s)". Read the decision field.
- An assertion that cannot fail is worse than none.
if before >= afterpasses exactly when the knob under test did nothing.
Portability
Windows runs go test ./internal/... in CI, and two habits break there:
exec.LookPathneeds an executable extension, so an extensionless shell-script stub is invisible. Skip on Windows like the existing stubs do, or use a real binary.os.Chmod(dir, 0o000)does not deny reads on Windows, so a test that provokes an unreadable-directory branch that way never enters it.- Never use a rooted literal such as
/repoas a test repository root: macOS and Linux call it absolute and Windows does not.
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.
- yesterday First seen · 75 lines · 738 tokens per session scan A 54634f572c33
procoder tests.instructions.md is an instructions file published in the GitHub repository azrtydxb/procoder (196 stars, last pushed 2d ago), licensed Apache-2.0. It adds 738 tokens to every session, about $0.0037 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 2026-08-30.
Other instructions, from other repositories
sdlc-quality AGENTS.md
Instructions for zircote-plugins/sdlc-quality, covering agents.md - openai codex guidelines, project identity, quick context, agent guidelines and 1. understand the domain model.
sdlc-quality copilot-instructions.md
Instructions for zircote-plugins/sdlc-quality, covering copilot instructions for sdlc plugin, project overview, repository structure, key conventions and rfc 2119 terminology.
high-quality-projects-skill AGENTS.md
AGENTS.md instructions for RandyNorthrup/high-quality-projects-skill, covering agents.md, the two workflows, resolving paths, what is in here and non-negotiables when using this package.
code-discipline-skills AGENTS.md
Instructions for eliranpv11/code-discipline-skills, covering code discipline, priority order, 1. think before coding, 2. verify reality before acting and 3. simplicity first.
high-quality-projects-skill copilot-instructions.md
Copilot instructions for RandyNorthrup/high-quality-projects-skill: This repository provides two vendor-neutral workflows for holding a codebase to production standards.
arbiterx CLAUDE.md
Instructions for NeelPrime/arbiterx, covering arbiterx engineering rules, setup, rules (always enforced), never generate and quality gate.