GFramework/.agents/skills/gframework-boot/SKILL.md

name, description

name

description

gframework-boot

Repository-specific boot workflow for the GFramework repo. Use when Codex needs to start or resume work in this repository from short prompts such as "boot", "continue", "read AGENTS", or "start the next step"; when the user expects Codex to first read AGENTS.md, .ai/environment/tools.ai.yaml, and public ai-plan tracking files; or when Codex should assess task complexity, decide whether explorer or worker subagents are warranted, and then proceed under the repository's workflow rules.

GFramework Boot

Overview

Use this skill to bootstrap work in the GFramework repository with minimal user prompting. Treat AGENTS.md as the source of truth. Use this skill to enforce a startup sequence, not to replace repository rules. If the task clearly requires the main agent to keep coordinating multiple parallel subagents while maintaining ai-plan and reviewing each result, switch to gframework-multi-agent-batch after the boot context is established.

Startup Workflow

1. Read AGENTS.md before choosing tools, planning edits, or delegating work.
2. Read .ai/environment/tools.ai.yaml to confirm the preferred local toolchain.
3. Read ai-plan/public/README.md before asking the user for missing context.
4. If ai-plan/public/README.md maps the current branch or worktree to active topics, inspect those topics' todos/ and traces/ directories in listed priority order.
5. If no mapping exists, scan ai-plan/public/<topic>/todos/ and ai-plan/public/<topic>/traces/ across active topics, and ignore ai-plan/public/archive/ unless the user explicitly asks for historical context.
6. Treat ai-plan/public/<topic>/archive/ as secondary context even for active topics; only read it when the active todo/trace files point there or when the user explicitly asks for historical detail.
7. If ai-plan/private/<branch-or-worktree>/ exists and is relevant, treat it as private recovery context for the current worktree only and do not assume it should be committed.
8. Classify the task state:
   • new: no matching recovery document exists, or the user is clearly starting fresh work
   • resume: a matching todo or trace exists and the user is continuing that thread
   • recovery: prior work looks partial, interrupted, or ambiguous and the next safe recovery point must be reconstructed
9. Choose the best matching ai-plan artifacts:
   • Prefer topics explicitly mapped from ai-plan/public/README.md
   • Prefer path names or headings that match the user's task wording
   • Break ties by most recently updated trace or todo
   • If ambiguity would materially change implementation, summarize the candidates and ask one concise question
10. Classify the task complexity before deciding on subagents:

• simple: one concern, one file or module, no parallel discovery required
• medium: a small number of modules, some read-only exploration helpful, critical path still easy to keep local
• complex: cross-module design, migration, large refactor, or work likely to exceed one context window

11. Estimate the current context-budget posture before substantive execution:

• account for loaded startup artifacts, active ai-plan files, visible diffs, open validation output, and likely next-step output volume
• if the task already appears near roughly 80% of a safe working-context budget, prefer closing the current batch, refreshing recovery artifacts, and stopping at the next natural semantic boundary instead of starting a fresh broad slice

12. Apply the delegation policy from AGENTS.md:

• Keep the critical path local
• Use explorer with gpt-5.1-codex-mini for narrow read-only questions, tracing, inventory, and comparisons
• Use worker with gpt-5.4 only for bounded implementation tasks with explicit ownership
• Do not delegate purely for ceremony; delegate only when it materially shortens the task or controls context growth
• If the user explicitly wants the main agent to keep orchestrating multiple workers through several review/integration cycles, prefer gframework-multi-agent-batch over ad-hoc delegation

13. Before editing files, tell the user what you read, how you classified the task, whether subagents will be used, and the first implementation step.
14. Proceed with execution, validation, and documentation updates required by AGENTS.md.

Task Tracking

For multi-step, cross-module, or interruption-prone work, maintain the repository recovery artifacts instead of keeping state only in chat.

• Update ai-plan/public/README.md whenever the active topic set or worktree mapping changes.
• Update the active public document under ai-plan/public/<topic>/todos/ with completed work, validation results, risks, and the next recovery point.
• Update the matching public trace under ai-plan/public/<topic>/traces/ with key decisions, delegated scope, and the immediate next step.
• Keep the active todo/trace files concise enough for boot to use as default entrypoints. When completed, validated stages start piling up, move their detailed history into ai-plan/public/<topic>/archive/ and leave archive pointers in the active files.
• Move stage-complete artifacts into ai-plan/public/<topic>/archive/, and move completed topics into ai-plan/public/archive/<topic>/ so boot does not keep reloading stale context.
• Keep worktree-private scratch recovery files under ai-plan/private/ and do not treat them as commit targets.
• Never write secrets, machine-specific paths, or other sensitive environment details into any ai-plan/** artifact.
• If the task is clearly complex and no recovery artifact exists yet, create one before substantive edits.

Recovery Heuristics

• If the user says next step, continue, 继续, or similar resume language, read ai-plan/public/README.md first, then search the mapped active topics before scanning the broader public area.
• If the current branch and the mapped active topics describe the same feature area, prefer resuming those topics first.
• If the repository state suggests in-flight work but no recovery document matches, reconstruct the safest next step from code, tests, and Git state before asking the user for clarification.
• If the current turn already carries heavy recovery context, broad diffs, or long validation output, prefer a recovery-point update and a clean stop over starting another large slice just because the code task itself remains open.

Example Triggers

• boot
• Use $gframework-boot and continue the current task
• Read AGENTS and public ai-plan, then start the next step
• 继续当前任务，先看 AGENTS.md 和 public ai-plan

References

Read references/startup-artifacts.md when you need a quick reminder of the repository entrypoints, task-state heuristics, or delegation defaults without re-reading the entire skill.