Skip to content

Yuimi-chaya/codex-development-guidelines / skills/adopt-agent-policy

adopt-agent-policy

Verified

Configure Codex user-level AGENTS.md when the user asks to install, adopt, or update these guidelines. Inspect the actual environment, relevant Codex configuration, and exposed models first; research public model facts, recommend a configuration, collect unresolved preferences together, and apply the authorized changes. Honor an explicit project-level target. Reviewing the repository or ordinary coding does not request installation.

references/profile.md

View on GitHub
# Optional Adoption Record

This is a small, optional provenance record for later updates. It is not an executable profile, a hidden loader, or a competing instruction authority. Effective preferences must be written into the approved target instructions; future coding threads should work without this record or the source repository.

## Contents and ownership

Do not create this record by default or ask about it as a prerequisite. If requested, resolve its location, visibility, and write scope in the same configuration proposal. It may be private or repository-owned. Do not persist a transcript, credentials, private network details, unnecessary hardware inventories, or unapproved model mappings.

Useful fields:

| Field | Meaning |
|---|---|
| Scope and target | Exact user/project scope and effective instruction file |
| Source | Selected repository/path and commit/release; mark uncommitted draft explicitly |
| Language | Runtime language and any separately maintained companion |
| Modules | Adopted, conditional, or omitted, with reasons |
| Decisions | Value, provenance, scope, and whether explicit or proposed-and-approved; maintenance IDs may be used here when useful, never required in runtime instructions |
| Deferred values | Why unknown; exact adopted conditional fallback, without implied permission |
| Model research, if requested | Exact exposed identity/provider, evidence sources and check date, uncertainty, approved mapping/fallback; distinguish research from selection |
| Interactive execution | Approved tool/model/context choices, direct-use authority, progress/stop budget, supported media/retention controls and unknowns |
| User interaction | Proactive-question/assistance preference, supported channel and fallback, source of choice, separately authorized feature writes, pending reload or verification |
| Environment decisions | Detected shell/encoding/Git facts, accepted or declined recommendations, retained-environment recovery approach |
| Recovery | Backup location, base identity, restore scope, verified integrity |
| Applied state | Exact approved changes, target hash, validation, pending human acceptance |

## Example

An illustrative record might state:

```text
Scope: project
Target: AGENTS.md
Source: user-selected local policy checkout, reviewed commit recorded at adoption
Language: Simplified Chinese; explicit in the adoption request
Delegation: disabled; existing project instruction, unchanged
Browser route: structured tool where allowed; otherwise ask for a scoped exception
Media numeric budget: deferred; inspect size, keep originals, send small review copies
Writes: policy diff approved; Skill installation not requested
```

The example is not a default profile. In particular, disabling delegation must not silently enable the primary Agent's interactive browser use.

## Updates

Reconcile the record with current instructions and facts before using it. If it is missing, continue from the actual target and available source; do not replay adoption merely to reconstruct bookkeeping. If it conflicts, treat it as a lead to investigate, not authority to overwrite the user's policy.

Research observations can become stale without invalidating every adopted rule. Refresh only affected evidence and decisions when requested or necessary; do not copy volatile price tables into runtime instructions or rerank models at every thread start. Never save credentials or private deployment identifiers as public research provenance.
reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 471kUpdated Claude CodeCodexVerifiedInstall
diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness.Stars: 322kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 322kUpdated Claude CodeCodexVerifiedInstall
executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is availableStars: 322kUpdated Claude CodeCodexVerifiedInstall