CtrlK
BlogDocsLog inGet started
Tessl Logo

oh-my-ai/agent-init

Creates, updates, or prunes an AGENTS.md for any repository by auditing the codebase, detecting non-discoverable gaps, and drafting minimal high-signal instructions that agents cannot infer from reading the code.

90

1.06x
Quality

94%

Does it follow best practices?

Impact

78%

1.06x

Average score across 3 eval scenarios

SecuritybySnyk

Passed

No known issues

Overview
Quality
Evals
Security
Files

criteria.jsonevals/scenario-3/

{
  "context": "Tests whether the agent correctly recommends hierarchical AGENTS.md files for a large monorepo, avoids asking generic discoverable questions while asking targeted questions about genuine ambiguities, and keeps questions limited to the prescribed maximum.",
  "type": "weighted_checklist",
  "checklist": [
    {
      "name": "Hierarchical structure recommended",
      "description": "The root AGENTS.md includes a '## Scope & routing' section that identifies which packages have their own AGENTS.md files, OR the agent creates package-level AGENTS.md files in at least 2 of the package directories",
      "max_score": 12
    },
    {
      "name": "No monolithic file for all modules",
      "description": "The root AGENTS.md does NOT contain detailed per-package guidance for all 6 packages in a single file without routing to package-level files",
      "max_score": 10
    },
    {
      "name": "No generic discoverable questions",
      "description": "The questions file or plan does NOT include questions about the tech stack, programming language, testing framework, or other information discoverable from reading the repo files",
      "max_score": 10
    },
    {
      "name": "Targeted ambiguity question asked",
      "description": "The agent produces a questions.md (or similar file) with at most 3 questions, each tied to a specific genuine ambiguity found in the repo (e.g., the conflicting deploy scripts for the payments package, or the meaning of the DO-NOT-DEPLOY marker)",
      "max_score": 12
    },
    {
      "name": "Question count within limit",
      "description": "No more than 3 questions are posed to the user — the questions.md or equivalent file contains at most 3 items",
      "max_score": 8
    },
    {
      "name": "CI vs local difference documented",
      "description": "AGENTS.md (root or package-level) documents the non-obvious 'INTEGRATION=true' environment variable required by CI but not present in local test scripts or .env.example",
      "max_score": 10
    },
    {
      "name": "Do-not-touch constraint documented",
      "description": "AGENTS.md documents the constraint on packages/payments/ (marked DO-NOT-DEPLOY, requires SRE sign-off) — not just that the directory exists",
      "max_score": 10
    },
    {
      "name": "No tech stack summaries in output",
      "description": "AGENTS.md files produced do NOT contain sections describing the tech stack (language, framework, package manager) as general context",
      "max_score": 8
    },
    {
      "name": "Scope statement present",
      "description": "The root AGENTS.md has a one-line or two-sentence scope statement immediately after the '# Agent instructions' heading",
      "max_score": 8
    },
    {
      "name": "Audit trail created",
      "description": "An audit-notes.md file exists with at least 3 entries documenting items considered but excluded, each citing the file that makes it discoverable",
      "max_score": 12
    }
  ]
}

evals

SKILL.md

tile.json