feat: add UserPromptSubmit hook for context-related skill activation

[Nathan Schram (nathanschram)]

Summary

• Adds context-forced-eval.sh — a new UserPromptSubmit hook that keyword-gates context-related prompts and injects skill evaluation instructions, improving NL trigger activation for ContextDocs skills
• Updates Context Guard install/uninstall/status to include the new hook and UserPromptSubmit event
• Adds CODEOWNERS, .gitignore for dev artifacts, and _typos.toml entry for hook regex patterns
• Full test coverage: 51 new test cases (Section 7) covering slash commands, NL keywords, non-matching prompts, Untether gating, and case insensitivity — total now 116/116

Files changed

File	Change
hooks/context-forced-eval.sh	New hook script (keyword-gated UserPromptSubmit)
.claude/hooks/context-forced-eval.sh	Installed copy
.claude/skills/context-guard/SKILL.md	Documents new hook + UserPromptSubmit settings
AGENTS.md	Hook count 6→7, lists new hook
CHANGELOG.md	Unreleased section with 3 fixes
commands/context-guard.md	Install/uninstall/status updated
llms.txt	New hook entry
scripts/update-installed-hooks.sh	Deploys new hook + patches settings.json
tests/test-hooks.sh	51 new tests (Section 7)
.github/CODEOWNERS	Repository ownership
.gitignore	Excludes dev artifacts
_typos.toml	Allows promot partial match

Test plan

• bash tests/test-hooks.sh — 116/116 pass
• bash tests/check-token-budgets.sh — 0 warnings
• bash tests/validate-llms-txt.sh — 0 errors

🤖 Generated with Claude Code

Summary by CodeRabbit

• New Features

  • Added automatic evaluation of user prompts for context-related keywords with intelligent skill recommendations.

• Bug Fixes

  • Fixed false positives in context detection for infrastructure files.
  • Improved context skill trigger reliability for stale file scenarios.
  • Synchronized skill descriptions across all documentation.

• Tests

  • Added comprehensive test coverage for context evaluation feature.

[coderabbitai]

Caution

Review failed

The pull request is closed.

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro

Run ID: 4904b60f-317b-43af-a9c3-2e235400d63a

📥 Commits

Reviewing files that changed from the base of the PR and between 0cb4ed1 and 1edd664.

📒 Files selected for processing (13)

• .claude/hooks/context-forced-eval.sh
• .claude/skills/ai-context/SKILL.md
• .claude/skills/context-guard/SKILL.md
• .github/CODEOWNERS
• .gitignore
• AGENTS.md
• CHANGELOG.md
• _typos.toml
• commands/context-guard.md
• hooks/context-forced-eval.sh
• llms.txt
• scripts/update-installed-hooks.sh
• tests/test-hooks.sh

---

📝 Walkthrough

Walkthrough

A new UserPromptSubmit hook is introduced that activates on user prompts to detect context-related keywords and inject skill evaluation context. Hook registration, installation logic, tests, and documentation are updated across the codebase. Skill descriptions clarify scope for ai-context and context-guard.

Changes

Cohort / File(s)	Summary
Hook Script Implementation .claude/hooks/context-forced-eval.sh, hooks/context-forced-eval.sh	New Bash hook script that detects context keywords via exact slash-command prefixes (/ai-context, /context-guard, /context-verify) or regex patterns. Outputs JSON with skill evaluation context if matched, empty object otherwise. Early-exits in Untether sessions.
Installation & Configuration scripts/update-installed-hooks.sh, commands/context-guard.md	Updates installation scripts to register context-forced-eval.sh in Tier 1 hooks and add UserPromptSubmit entry to settings.json. Adds detection logic and warnings for missing UserPromptSubmit hook during installation/uninstall flows.
Skill Descriptions .claude/skills/ai-context/SKILL.md, .claude/skills/context-guard/SKILL.md	Expands ai-context skill description to clarify scope: added emphasis on updating/fixing out-of-date context, refreshing files, and promoting patterns. Adds context-guard documentation for the new UserPromptSubmit hook integration.
Project Documentation AGENTS.md, CHANGELOG.md, llms.txt	Updates hook count from 6 to 7 in AGENTS.md. Adds Context Forced Eval Hook entry to llms.txt manifest. CHANGELOG documents three documentation-related fixes including improved skill descriptions and context guard reliability.
Environment & Metadata .github/CODEOWNERS, .gitignore, _typos.toml	Adds CODEOWNERS entry routing all paths to core team. Expands .gitignore with three new directories (ai-context-workspace/, incoming/, tests/activation-results/). Adds regex partial match entry promot for promotion-related spell checking.
Test Coverage tests/test-hooks.sh	Adds new Section 7 test suite for context-forced-eval.sh covering slash-commands, NL keywords, non-matching prompts, gate logic, and cleanup. Includes duplicate test section (appears twice). Updates header comment to reference 7 ContextDocs hooks.

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~20 minutes

Possibly related PRs

• fix: Context Guard hooks false positive on infrastructure rule files #13: Modifies ai-context SKILL.md description and adjusts Context Guard hook exclusions, directly overlapping with this PR's skill documentation and hook functionality changes.

Poem

🐰 With twitching nose and keywords keen,
Our hook-eval hops the prompt scene,
Context detected, skills aligned—
The finest guidance, smartly designed! ✨

✨ Finishing Touches

• 📝 Generate docstrings (stacked PR)
• 📝 Generate docstrings (commit on current branch)

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Post copyable unit tests in a comment
• Commit unit tests in branch fix/ai-context-nl-triggering

📝 Coding Plan

• Generate coding plan for human review comments

---

_{Comment @coderabbitai help to get the list of available commands and usage tips.}