Skip to content

onboarding-checklist.md

Latest commit

 

History

History
141 lines (106 loc) · 5.92 KB

onboarding-checklist.md

File metadata and controls

141 lines (106 loc) · 5.92 KB

Outcome Checklist

Use this file as the canonical progress tracker during onboarding. The goal is not to mark steps complete; it is to prove that the repository can support a small set of agent capabilities with real evidence.

This file is intended to be copied into the target repository and updated there throughout onboarding.

How To Use

  1. Start with Validate Current State so the repository has a baseline.
  2. Work toward one proof at a time. Do not try to unlock every outcome at once.
  3. Record evidence each time an outcome changes state.
  4. Keep .agents/code-mint-status.json in sync with this checklist. It is the machine-readable index; this file is the detailed evidence record.
  5. Revisit the file when the user comes back later. It should be obvious what is done, blocked, or still unproven.

Status Key

Status Meaning
Not Started No meaningful work has been done on this outcome yet.
Blocked The next proof is known, but something external or risky is preventing it.
In Progress Work is underway, but the outcome is not yet proven with evidence.
Proven The outcome has been demonstrated and the proof is captured below.
N/A The outcome does not apply to this repository.

Outcome Tracker

Outcome Status Last Validated
Validate Current State Not Started —
Navigate Not Started —
Self-Test Not Started —
Smoke Path Not Started —
Bug Reproduction Not Started —
SRE Investigation Not Started —

Outcome Details

Copy one section per outcome as it moves forward. Each section is the detailed evidence record for that outcome.

Validate Current State

The agent can assess the repo before making changes.

  • Status: Not Started
  • Date: —
  • Evidence: applicable audit reports plus .agents/reports/onboarding-summary.md
  • Exact Check: Run all applicable Phase 1 auditors and summarize what is working, blocked, risky, and next to prove.
  • What Passed: —
  • What Is Still Missing: —
  • Next Action: Establish this first.

Navigate

The agent can explain where to work and why modules exist.

  • Status: Not Started
  • Date: —
  • Evidence: AGENTS.md, subdirectory AGENTS.md, sample repo explanation
  • Exact Check: Ask the agent to explain the repo map and target module for a sample task, then record the grounded answer.
  • What Passed: —
  • What Is Still Missing: —
  • Next Action: Favor durable repo-local docs over chat-only explanations.

Self-Test

The agent can run the smallest relevant automated checks and trust the result.

  • Status: Not Started
  • Date: —
  • Evidence: exact test command, test file, CI reference
  • Exact Check: Run one targeted test path for a real module or behavior change and capture the pass/fail signal.
  • What Passed: —
  • What Is Still Missing: —
  • Next Action: Record what is still manual or flaky.

Smoke Path

The agent can prove the runtime is meaningfully usable.

  • Status: Not Started
  • Date: —
  • Evidence: runtime guide, screenshot, log excerpt, curl output
  • Exact Check: Execute one trusted, non-destructive smoke path with documented prerequisites, steps, stop conditions, and success signal.
  • What Passed: —
  • What Is Still Missing: —
  • Next Action: Smallest safe check wins.

Bug Reproduction

The agent can reproduce a reported issue before proposing a fix.

  • Status: Not Started
  • Date: —
  • Evidence: failing test, script, or repro recipe linked to the issue
  • Exact Check: Turn one real bug report into a deterministic failing case another person or agent can rerun.
  • What Passed: —
  • What Is Still Missing: —
  • Next Action: The repro is the proof.

SRE Investigation

The agent can inspect logs, metrics, traces, CI, or infra evidence.

  • Status: Not Started
  • Date: —
  • Evidence: investigation note, query output, linked dashboard/log command
  • Exact Check: Gather operational evidence, produce a ranked hypothesis, and record the next actions or N/A reason.
  • What Passed: —
  • What Is Still Missing: —
  • Next Action: Mark N/A only when no operational tooling exists.

Suggested Evidence Patterns

Outcome Good Evidence
Validate Current State Audit reports plus a short plain-language summary in .agents/reports/onboarding-summary.md covering working, blocked, risky, and next proof
Navigate Root and module AGENTS.md files plus a sample repo explanation that answers where a representative task should happen
Self-Test Exact targeted test command, the module or behavior it verifies, and the pass/fail signal it produced
Smoke Path Prerequisites, startup order, smoke command, concrete success signal, and "stop here" safety notes
Bug Reproduction Failing test or deterministic repro script tied to a reported issue and rerunnable by another person or agent
SRE Investigation Query commands, time range, observed symptoms, ranked hypotheses, and next actions or an N/A reason

Skill Mapping

Outcome Skills To Reach It
Validate Current State meta--onboarding plus all applicable auditor skills
Navigate legibility--auditor, legibility--enhancer
Self-Test autonomy--test-readiness-auditor, autonomy--test-readiness-creator
Smoke Path autonomy--env-auditor, autonomy--env-creator, autonomy--runtime-auditor, autonomy--runtime-creator
Bug Reproduction autonomy--sre-agent, autonomy--test-readiness-*
SRE Investigation autonomy--sre-auditor, autonomy--sre-agent

Return-Visit Questions

When someone comes back to onboarding later, this file should answer:

  • What has been proven already?
  • What is the next missing proof?
  • Which command, prompt, or scenario should be run next?
  • What is blocked on human approval or missing tooling?