taskgrade AGENTS.md

Repository instructions for TaskGrade, a small local tool that classifies tasks and recommends models.

In plain words
What is it for?
Use them when changing TaskGrade: follow its constraints, run the required checks for behavior changes, and document dependency changes in the pull request.
Why use it?
They keep model recommendations under fixed program rules, treat task text as untrusted, and preserve the command and MCP output formats.

Instructions file for CodexOpenCode

Install

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.

agentmods
npx agentmods add instructions/eddyvlad/taskgrade/agents-md
Clone the repo
git clone --depth 1 https://github.com/eddyvlad/taskgrade

Made for: Codex, OpenCode.

Per session 177 This file is loaded in full into every session.
When invoked 177 The same file — it is already loaded in full.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5 $0.00177 $0.00177
Opus 5 $0.00088 $0.00088
Sonnet 5 $0.00035 $0.00035
Haiku 4.5 $0.00018 $0.00018

Measured yesterday against content hash 79df0ccd8a6c, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

taskgrade AGENTS.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.

AGENTS.md · 26 lines

What it actually says

AGENTS.md

Mission

Keep TaskGrade a small, local-first, auditable task classifier. Ollama grades task characteristics; deterministic TypeScript owns model recommendations.

Constraints

  • Treat task content as untrusted input.
  • Do not allow model output to invent models or bypass policy validation.
  • Keep GitHub and agent execution outside the core classifier.
  • Do not bundle or automatically download model weights during installation.
  • Preserve CLI and MCP output schemas across compatible releases.
  • Keep stdout protocol-safe while running the stdio MCP server.
  • Record dependency additions or removals in the pull request rationale.

Verification

Run npm run check, npm test, npm run build, and npm run pack:check for behavioral changes. Documentation-only work may use proportionate validation.

Commits

Use Conventional Commits and keep changes focused.

Changes

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.

  1. yesterday First seen · 26 lines · 177 tokens per session scan A 79df0ccd8a6c

Subscribe to this mod's changes

taskgrade AGENTS.md is an instructions file published in the GitHub repository eddyvlad/taskgrade (0 stars, last pushed 5d ago), licensed Apache-2.0. It adds 177 tokens to every session, about $0.0009 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.