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/raydestar/sir-thaddeus/e2e-harness-rulesgit clone --depth 1 https://github.com/raydeStar/sir-thaddeusWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/instructions/raydestar/sir-thaddeus/e2e-harness-rules)<a href="https://agentmods.dev/instructions/raydestar/sir-thaddeus/e2e-harness-rules"><img src="https://agentmods.dev/badge/instructions/raydestar/sir-thaddeus/e2e-harness-rules.svg" alt="Measured on agentmods" height="20"></a>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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.02007 | $0.02007 |
| Opus 5 | $0.01004 | $0.01004 |
| Sonnet 5 | $0.00401 | $0.00401 |
| Haiku 4.5 | $0.00201 | $0.00201 |
Grade A, and why
E2E Harness Self-Fixing Protocol 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 6d ago.
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 — 182 lines — stays where its author put it; the contents beside it link to each section on GitHub.
E2E Harness — Self-Fixing Protocol
You are working on a test harness that evaluates AI response quality. Your job is to fix the underlying agent logic so it produces better responses. You are NOT trying to make tests pass. You are trying to make the agent work correctly.
Running the Harness
# Full run — all suites
.\dev\harness.ps1 --all --judge none
# Single suite
.\dev\harness.ps1 --suite <name> --judge none
# Single test within a suite
.\dev\harness.ps1 --suite <name> --test <test_id> --judge none
# With AI judge scoring (requires local LLM or --judge cursor)
.\dev\harness.ps1 --suite <name> --judge model
The harness project lives at tools/SirThaddeus.Harness/. The entry point script is .\dev\harness.ps1.
Reading Failure Artifacts
Every test run writes artifacts to:
artifacts/harness/<run-id>/<suite>/<test-id>/iter-XX/
├── input.json ← prompt sent to the agent
├── steps.jsonl ← full tool call and result trace
├── final.txt ← the agent's final response text
├── score.json ← detailed score breakdown with penalties
├── judge_packet.json ← context sent to judge (if enabled)
└── judge_result.json ← judge verdict (if enabled)
When diagnosing a failure, read these files in this order:
score.json— which scoring dimensions took the hit?final.txt— is this a real answer or a deflection?steps.jsonl— did the agent call tools? did it use the results?input.json— what was the agent actually asked?
If score.json says PASS but final.txt is clearly weak, hedged, incorrect, or unsupported, continue diagnosis. Treat the run as suspicious until the discrepancy is explained.
When reading steps.jsonl, break the run into explicit stages. For each tool call, note:
- the exact query or arguments
- what came back
- whether the result was relevant
- what the agent did next because of that result
If there were multiple searches, retries, or fallbacks, document them separately instead of collapsing them into "the agent searched".
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.
- 6d ago First seen · 182 lines · 2,007 tokens per session scan A 4dd44293f99b
E2E Harness Self-Fixing Protocol is an instructions file published in the GitHub repository raydeStar/sir-thaddeus (13 stars, last pushed 5d ago), licensed Apache-2.0. It adds 2,007 tokens to every session, about $0.0100 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
llm-swarm-router AGENTS.md
AGENTS.md instructions for matthewdcage/llm-swarm-router, covering agent & developer guide, dox rail, project overview, architecture and repository layout.
llm-swarm-router CLAUDE.md
Claude Code instructions for matthewdcage/llm-swarm-router, covering claude code: swarm-llm and honcho memory (mcp).
agent-audit AGENTS.md
AGENTS.md instructions for fullstackdegen/agent-audit, covering agent audit instructions, core rules, applying fix packs, what this tool covers and completion standard.
mcp-local-llm AGENTS.md
Instructions for aplaceforallmystuff/mcp-local-llm, covering agents.md - mcp-local-llm, tech stack, architecture, development commands and environment variables.
mcp-explorer AGENTS.md
AGENTS.md instructions for MausRundung/mcp-explorer, covering agents.md, start of every chat, tools, listalloweddirectories and exploreproject.
idesense CLAUDE.md
Instructions for vcth4nh/idesense, covering idesense, agent rules and where things live.