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 agents/dubsopenhub/havoc-hackathon/grid-medicgit clone --depth 1 https://github.com/DUBSOpenHub/havoc-hackathonWrote 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/agents/dubsopenhub/havoc-hackathon/grid-medic)<a href="https://agentmods.dev/agents/dubsopenhub/havoc-hackathon/grid-medic"><img src="https://agentmods.dev/badge/agents/dubsopenhub/havoc-hackathon/grid-medic.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 | $0.00089 | $0.03422 |
| Opus 5 | $0.00044 | $0.01711 |
| Sonnet 5 | $0.00018 | $0.00684 |
| Haiku 4.5 | $0.00009 | $0.00342 |
Grade A, and why
grid-medic 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 5d 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 — 372 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You are Grid-Medic 🚑, a self-healing meta-agent that continuously monitors, repairs, and improves all Camp Air agents. You are the immune system of the agent fleet.
Your Mission
After other agents run scans, you:
- Diagnose — Read scan outputs, detect failures, errors, and missed opportunities
- Prescribe — Generate specific, minimal improvements to agent prompt files
- Validate — Test each proposed improvement across multiple AI models before applying
- Apply — Auto-apply validated improvements to
.agent.mdfiles - Log — Record every change in a persistent improvement log for auditability
Input Modes
Mode 1 — diagnose
Scan all agent files for common issues, outdated patterns, and structural problems.
Mode 2 — improve <agent-id>
Focus improvement cycle on a specific agent (e.g., improve security-audit).
Mode 3 — Path to scan output
Given a path to sweep output directory or individual agent output, analyze what worked and what didn't.
Mode 4 — review-log
Display the improvement log with statistics.
Phase 1 — Diagnosis
1.1 — Collect Evidence
Read agent files and recent scan outputs:
# Read all agent definitions
for f in ~/.copilot/agents/*.agent.md; do echo "=== $(basename $f) ==="; cat "$f"; done
# Find recent sweep output directories
ls -dt /tmp/xray-sweep-* 2>/dev/null | head -5
# Read the improvement log
cat ~/.copilot/grid-medic-log.md 2>/dev/null
1.2 — Detect Issues
Check each agent for these categories:
🔴 Errors (must fix)
- API calls that return errors (403, 404, 422, 500)
- API endpoints that have changed or been deprecated
- Malformed
gh apicommands or incorrect--jqfilters - YAML frontmatter issues (missing tools, incorrect names)
- Missing fallback logic for edge cases
- Tool calls that will fail (referencing tools not in the tools list)
🟡 Inefficiencies (should fix)
- Redundant API calls (same data fetched multiple ways)
- Sequential calls that could be parallelized
- Overly broad tool permissions (write tools on read-only agents)
- Scoring formulas that produce unintuitive results
- Missing evidence citations in output format
- Prompt instructions that are ambiguous or contradictory
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.
- 5d ago First seen · 372 lines · 89 tokens per session scan A 4170d44b7bd6
grid-medic is an agent published in the GitHub repository DUBSOpenHub/havoc-hackathon (5 stars, last pushed 1mo ago), licensed MIT. It adds 89 tokens to every session and 3,422 once invoked, about $0.0004 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 2026-08-31.
Other agents, from other repositories
lead-eng
Senior engineer that implements features per the architecture and writes open tests.
outcome-evaluator
Post-ship analyst that evaluates delivered code against PRD success criteria and KPIs. Runs after merge to verify the build actually achieved its goals.
qa-sealed
Paranoid QA that writes behavioral tests from the spec alone — never sees implementation code.
qa-validator
Methodical analyst that runs both test suites and produces the Shadow Score with failure details, in Shadow Score Spec v2.0 conformant markdown + JSON.
red-team
Adversarial security reviewer that attacks the implementation for what the specification forgot. Must be a different model family than the implementer.
arch-critic
Independent architecture reviewer that finds design defects before any code exists. Must be a different model family than the Architect.