feat(hooks): add built-in prompt guards for branch protection and dirty tree by RyderFreeman4Logos · Pull Request #134 · RyderFreeman4Logos/cli-sub-agent

RyderFreeman4Logos · 2026-02-17T04:16:16Z

Summary

Add two default prompt guard scripts that ship with CSA out of the box:
- branch-protection: Warns when working on protected branches (main/master/dev/develop/release/*)
- dirty-tree-reminder: Reminds about uncommitted changes with file count
Built-in guards are active by default, disabled with builtin_guards = false in hooks.toml
User-defined [[prompt_guard]] entries replace built-ins entirely

Test plan

test_builtin_prompt_guards_returns_two_entries — verifies count, names, timeouts
test_builtin_branch_protection_on_feature_branch — no warning on non-protected branch
test_builtin_dirty_tree_reminder_on_clean_tree — guard executes without error
test_builtin_guards_loaded_by_default — config loads 2 guards with no user config
test_builtin_guards_replaced_by_user_config — user guards replace built-ins
test_builtin_guards_disabled_explicitly — builtin_guards = false disables all
just pre-commit passes (1123 tests, 0 failed)

🤖 Generated with Claude Code

RyderFreeman4Logos · 2026-02-17T04:16:25Z

@codex review

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 005a706869

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

chatgpt-codex-connector · 2026-02-17T04:19:31Z

crates/csa-hooks/src/guard_tests.rs

+        assert_eq!(
+            results.len(),
+            1,
+            "dirty-tree-reminder should produce exactly one result"


Make dirty-tree guard test tolerate clean worktrees

run_prompt_guards only includes guards that emit non-empty stdout, but this test requires one result even though dirty-tree-reminder intentionally prints nothing on a clean repository. In typical CI checkouts (clean tree), results.len() becomes 0, so this assertion makes the test fail despite correct behavior and can block the test suite.

Useful? React with 👍 / 👎.

Fixed in 1ba282f. Changed assertion to results.len() <= 1 — guard may produce 0 (clean tree) or 1 (dirty tree) results.

chatgpt-codex-connector · 2026-02-17T04:19:32Z

crates/csa-hooks/src/guard_tests.rs

+        // We are on feat/builtin-prompt-guards, so no warning expected
+        let guards = builtin_prompt_guards();
+        let branch_guard = &guards[0];
+        let results = run_prompt_guards(std::slice::from_ref(branch_guard), &test_context());
+        // On a feature branch, the guard should produce empty output


Remove branch-dependent assumption from branch guard test

This test assumes the current checkout is a feature branch, but it runs the built-in guard against whatever branch the test runner uses; on main/master/dev/develop/release/* the guard emits a warning and this assertion fails. Because no temporary repo/branch is created, the test result is environment-dependent and can become flaky across local and CI runs.

Useful? React with 👍 / 👎.

Fixed in 1ba282f. Test now only verifies execution succeeds (no panic/timeout), without asserting on output content.

RyderFreeman4Logos · 2026-02-17T04:21:13Z

@codex review

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 1ba282fa3f

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

chatgpt-codex-connector · 2026-02-17T04:24:36Z

crates/csa-hooks/src/guard.rs

+    vec![
+        PromptGuardEntry {
+            name: "branch-protection".to_string(),
+            command: r#"branch=$(git rev-parse --abbrev-ref HEAD 2>/dev/null) || exit 0


Run built-in git guards in the configured project root

The new built-in branch-protection (and the dirty-tree-reminder right below it) calls plain git ... without setting a working directory, but run_single_guard spawns sh -c without current_dir and determine_project_root supports --cd without changing process CWD, so running csa ... --cd <repo> from another directory makes these guards silently exit 0 (stderr is redirected to /dev/null) and skip protection exactly when users expect it. This regresses branch/dirty-tree safety for any workflow where project root differs from the launcher CWD.

Useful? React with 👍 / 👎.

Fixed in 8364aeb. Added current_dir(project_root) to run_single_guard's Command builder. All guard scripts now run in the configured project root directory, ensuring correct behavior with csa --cd <repo>.

RyderFreeman4Logos · 2026-02-17T04:27:09Z

@codex review

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 8364aeb2f0

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

chatgpt-codex-connector · 2026-02-17T04:31:55Z

crates/csa-hooks/src/config.rs

+        if !other.builtin_guards {
+            self.builtin_guards = false;


Let higher-priority config re-enable builtin guards

merge_with only updates builtin_guards when the incoming layer is false, so once any lower-priority config sets builtin_guards = false, later higher-priority layers can never restore the default guards (even if they explicitly set builtin_guards = true). This violates the stated layered-priority merge model and makes per-project override of a global disable impossible.

Useful? React with 👍 / 👎.

Fixed in 13cfdf6. Changed to self.builtin_guards = other.builtin_guards — higher-priority layer always wins, allowing project config to re-enable after global disable.

RyderFreeman4Logos · 2026-02-17T04:33:35Z

@codex review

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 13cfdf6dd1

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

chatgpt-codex-connector · 2026-02-17T04:38:02Z

crates/csa-hooks/src/config.rs

    /// (non-empty array wins; empty array means "no override from this layer").
    fn merge_with(&mut self, other: Self) {
+        // builtin_guards: higher-priority layer always wins (can re-enable after disable)
+        self.builtin_guards = other.builtin_guards;


Merge builtin_guards only when explicitly configured

builtin_guards is deserialized with #[serde(default = "default_true")], so any higher-priority hooks.toml that omits this key still carries true; this unconditional assignment means a project config with unrelated settings will silently re-enable built-in guards that were disabled in global config. In normal global+project loading, users cannot keep a global builtin_guards = false unless every higher layer repeats it, which breaks expected layered-override behavior for omitted fields.

Useful? React with 👍 / 👎.

Fixed in a2ebd5b. Changed builtin_guards to Option<bool>: None = not specified (inherits from lower layer), Some(true/false) = explicitly set. Added two new tests verifying global-disable+project-reenable and global-disable+project-omit-inherits behaviors.

RyderFreeman4Logos · 2026-02-17T04:41:37Z

@codex review

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: a2ebd5beee

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

chatgpt-codex-connector · 2026-02-17T04:47:39Z

crates/csa-hooks/src/guard.rs

+    vec![
+        PromptGuardEntry {
+            name: "branch-protection".to_string(),
+            command: r#"branch=$(git rev-parse --abbrev-ref HEAD 2>/dev/null) || exit 0


Use a branch detector that works before first commit

The built-in branch-protection guard uses git rev-parse --abbrev-ref HEAD and then || exit 0, which suppresses warnings when that command fails. In a newly initialized repo before the first commit (for example git init -b main), rev-parse exits non-zero (HEAD is unborn), so the guard emits nothing even on protected branches; this misses the reminder at the exact point users are likely to make an initial direct commit to main/master.

Useful? React with 👍 / 👎.

…ty tree reminder Add two default prompt guard scripts that ship with CSA out of the box: - branch-protection: Warns when working on protected branches (main/master/dev/ develop/release/*) to prevent accidental direct commits. - dirty-tree-reminder: Reminds about uncommitted changes before starting new work, showing the count of modified files. Built-in guards are active by default when no user-defined guards are configured. Users can disable them with `builtin_guards = false` in hooks.toml, or replace them entirely by defining their own `[[prompt_guard]]` entries. Configuration loading: built-in guards are injected as the lowest priority layer (Layer 0) after the 4-tier merge. Any user-defined prompt_guard array takes precedence. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

RyderFreeman4Logos · 2026-02-17T04:50:49Z

@codex review

chatgpt-codex-connector · 2026-02-17T04:56:13Z

Codex Review: Didn't find any major issues. Already looking forward to the next diff.

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

chatgpt-codex-connector bot reviewed Feb 17, 2026

View reviewed changes

RyderFreeman4Logos force-pushed the feat/builtin-prompt-guards branch from 005a706 to 1ba282f Compare February 17, 2026 04:20

chatgpt-codex-connector bot reviewed Feb 17, 2026

View reviewed changes

RyderFreeman4Logos force-pushed the feat/builtin-prompt-guards branch from 1ba282f to 8364aeb Compare February 17, 2026 04:26

chatgpt-codex-connector bot reviewed Feb 17, 2026

View reviewed changes

RyderFreeman4Logos force-pushed the feat/builtin-prompt-guards branch from 8364aeb to 13cfdf6 Compare February 17, 2026 04:33

chatgpt-codex-connector bot reviewed Feb 17, 2026

View reviewed changes

RyderFreeman4Logos force-pushed the feat/builtin-prompt-guards branch from 13cfdf6 to a2ebd5b Compare February 17, 2026 04:41

chatgpt-codex-connector bot reviewed Feb 17, 2026

View reviewed changes

RyderFreeman4Logos force-pushed the feat/builtin-prompt-guards branch from a2ebd5b to c281c8d Compare February 17, 2026 04:50

RyderFreeman4Logos merged commit d87be4a into main Feb 17, 2026
5 checks passed

RyderFreeman4Logos deleted the feat/builtin-prompt-guards branch February 17, 2026 04:57

Conversation

RyderFreeman4Logos commented Feb 17, 2026

Summary

Test plan

Uh oh!

RyderFreeman4Logos commented Feb 17, 2026

Uh oh!

chatgpt-codex-connector bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

chatgpt-codex-connector bot Feb 17, 2026

Choose a reason for hiding this comment

Uh oh!

RyderFreeman4Logos Feb 17, 2026

Choose a reason for hiding this comment

Uh oh!

chatgpt-codex-connector bot Feb 17, 2026

Choose a reason for hiding this comment

Uh oh!

RyderFreeman4Logos Feb 17, 2026

Choose a reason for hiding this comment

Uh oh!

RyderFreeman4Logos commented Feb 17, 2026

Uh oh!

chatgpt-codex-connector bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

chatgpt-codex-connector bot Feb 17, 2026

Choose a reason for hiding this comment

Uh oh!

RyderFreeman4Logos Feb 17, 2026

Choose a reason for hiding this comment

Uh oh!

RyderFreeman4Logos commented Feb 17, 2026

Uh oh!

chatgpt-codex-connector bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

chatgpt-codex-connector bot Feb 17, 2026

Choose a reason for hiding this comment

Uh oh!

RyderFreeman4Logos Feb 17, 2026

Choose a reason for hiding this comment

Uh oh!

RyderFreeman4Logos commented Feb 17, 2026

Uh oh!

chatgpt-codex-connector bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

chatgpt-codex-connector bot Feb 17, 2026

Choose a reason for hiding this comment

Uh oh!

RyderFreeman4Logos Feb 17, 2026

Choose a reason for hiding this comment

Uh oh!

RyderFreeman4Logos commented Feb 17, 2026

Uh oh!

chatgpt-codex-connector bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

chatgpt-codex-connector bot Feb 17, 2026

Choose a reason for hiding this comment

Uh oh!

RyderFreeman4Logos commented Feb 17, 2026

Uh oh!

chatgpt-codex-connector bot commented Feb 17, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!