| Dimension | Shared | Only in A | Only in B | Overlap |
|---|---|---|---|---|
| Sections | 4 | 1 | 1 | 67% |
| Commands | 0 | 0 | 0 | — |
| Section tags | 4 | 0 | 0 | 100% |
What each file covers
Sections
4 shared · 1 only in A · 1 only in B- − CLAUDE.md - SteadyAgent Claude Code Guide
- + AGENTS.md - SteadyAgent Codex Guide
- Project Intent
- Working Rules
- Validation
- Release Discipline
Commands
neither file has anySection tags
4 shared · 0 only in A · 0 only in B- code-style
- deployment
- do-not
- agent-behaviour
Line diff
Khalilzhang0825/steadyagent · CLAUDE.md
@@ −1 @@
1# CLAUDE.md - SteadyAgent Claude Code Guide
2
3This file gives Claude Code a compact project map for contributing to SteadyAgent.
4
5## Project Intent
6
7SteadyAgent is a local-first harness for AI coding agents. It turns ad-hoc Codex and Claude Code sessions into a repeatable loop with scoped planning, repository preflight, deterministic checks, lifecycle hooks, independent review, and checkpoint evidence.
8
9## Working Rules
10
11- Start from the relevant files and docs instead of exploring the whole repository.
12- Keep always-on guidance short; move detailed workflow knowledge into `rules/`, `skills/`, or `docs/`.
13- Prefer scripts and hooks for repeatable checks.
14- Preserve host differences honestly: Codex and Claude Code do not expose identical enforcement surfaces.
15- Keep public files free of private paths, local credentials, unpublished plans, and maintainer-only state.
16- Update English and Chinese documentation together when user-facing behavior changes.
17
18## Validation
19
@@ −36 @@
36## Release Discipline
37
38- Do not publish, retag, or rewrite remote history without explicit maintainer approval.
39- Treat each public release as an audited artifact: clean history, reproducible validation, and clear release notes.
40- Use [docs/github-publication-runbook.md](docs/github-publication-runbook.md) for public release steps.
41
Khalilzhang0825/steadyagent · AGENTS.md
@@ +1 @@
1# AGENTS.md - SteadyAgent Codex Guide
2
3This file gives Codex a compact project map for contributing to SteadyAgent.
4
5## Project Intent
6
7SteadyAgent is a local-first harness for AI coding agents. It packages concise instructions, progressive rules, validation scripts, lifecycle hook examples, and release evidence so agent-assisted development is easier to scope, verify, review, and recover.
8
9## Working Rules
10
11- Keep changes scoped to the requested workflow, tool, template, or documentation surface.
12- Read the nearest project guide, relevant docs, and directly touched scripts before editing.
13- Prefer deterministic scripts, tests, and hooks over longer instructions when behavior must be enforced.
14- Preserve the public release surface: no private paths, local credentials, unpublished plans, or maintainer-only notes.
15- Keep English as the default public surface; update the paired Chinese docs when user-facing behavior changes.
16- Use `rg` or `rg --files` for repository search.
17
18## Validation
19
@@ +36 @@
36## Release Discipline
37
38- Do not publish, retag, or rewrite remote history without explicit maintainer approval.
39- Treat `v1.0.0` and later releases as audited artifacts: clean history, reproducible validation, and clear release notes.
40- Use [docs/github-publication-runbook.md](docs/github-publication-runbook.md) for public release steps.
41
@@ −1 +1 @@
1−# CLAUDE.md - SteadyAgent Claude Code Guide
1+# AGENTS.md - SteadyAgent Codex Guide
22
3−This file gives Claude Code a compact project map for contributing to SteadyAgent.
3+This file gives Codex a compact project map for contributing to SteadyAgent.
44
55 ## Project Intent
66
7−SteadyAgent is a local-first harness for AI coding agents. It turns ad-hoc Codex and Claude Code sessions into a repeatable loop with scoped planning, repository preflight, deterministic checks, lifecycle hooks, independent review, and checkpoint evidence.
7+SteadyAgent is a local-first harness for AI coding agents. It packages concise instructions, progressive rules, validation scripts, lifecycle hook examples, and release evidence so agent-assisted development is easier to scope, verify, review, and recover.
88
99 ## Working Rules
1010
11−- Start from the relevant files and docs instead of exploring the whole repository.
12−- Keep always-on guidance short; move detailed workflow knowledge into `rules/`, `skills/`, or `docs/`.
13−- Prefer scripts and hooks for repeatable checks.
14−- Preserve host differences honestly: Codex and Claude Code do not expose identical enforcement surfaces.
15−- Keep public files free of private paths, local credentials, unpublished plans, and maintainer-only state.
16−- Update English and Chinese documentation together when user-facing behavior changes.
11+- Keep changes scoped to the requested workflow, tool, template, or documentation surface.
12+- Read the nearest project guide, relevant docs, and directly touched scripts before editing.
13+- Prefer deterministic scripts, tests, and hooks over longer instructions when behavior must be enforced.
14+- Preserve the public release surface: no private paths, local credentials, unpublished plans, or maintainer-only notes.
15+- Keep English as the default public surface; update the paired Chinese docs when user-facing behavior changes.
16+- Use `rg` or `rg --files` for repository search.
1717
1818 ## Validation
1919
@@ −36 +36 @@
3636 ## Release Discipline
3737
3838 - Do not publish, retag, or rewrite remote history without explicit maintainer approval.
39−- Treat each public release as an audited artifact: clean history, reproducible validation, and clear release notes.
39+- Treat `v1.0.0` and later releases as audited artifacts: clean history, reproducible validation, and clear release notes.
4040 - Use [docs/github-publication-runbook.md](docs/github-publication-runbook.md) for public release steps.
4141
