glaz-j/codex-freshread/freshread
Claude Code and agent skill by glaz-j/codex-freshread
- community
Checking that the SKILL.md file is still in the repository.
Inspect and manage FreshRead file-context freshness for the current Codex session. Use when the user asks for FreshRead status, inspection, clearing, stale-context diagnosis, or help after FreshRead blocks an edit. Do not use as a generic repository or Git change monitor.
Install
npx vskill@latest install glaz-j/codex-freshread/freshreadSKILL.md
FreshRead
FreshRead hooks run automatically for qualifying supported reads. They track file snapshots from those reads, compare the tracked Read Set with filesystem reality at lifecycle boundaries, and guard supported edits. A reported hook failure or timeout means the guard is unavailable until repaired.
When FreshRead reports STALE_SELF, STALE_EXTERNAL, or MISSING, re-read the named path from the filesystem before relying on remembered content. Retry a blocked edit only after the re-read establishes a new snapshot. Treat UNKNOWN as unknown; never describe it as fresh.
The user-facing commands are handled by the UserPromptSubmit hook:
freshread statusreturns state counts for this session.freshread inspectlists tracked paths and details; an optional path narrows the result.freshread clearclears only this session's Read Set. Clear state only when the user explicitly requests it.
Coverage is intentionally file-level and bounded. apply_patch targets are guarded exactly, and recognized structured or canonical success can be attributed as STALE_SELF. The default observed-read protocol does not rewrite commands: PreToolUse registers one literal-path, non-empty complete-file candidate, and PostToolUse records it only when the complete canonical raw string exactly matches the stable current file rendering. PowerShell reads require Get-Content/gc -Raw -Encoding utf8; when a canonical payload omits tool-level workdir/cwd, use a fully qualified path. Shell writes are guarded before execution, but a canonical payload without a shell hint is not rewritten or self-attributed; nonce-bound write evidence requires an explicit compatible shell hint. Search/partial reads, empty or truncated output, arbitrary scripts, opaque shell pipelines, MCP/hosted reads, specialized tool paths, and external processes are not reliable snapshots. Do not claim whole-workspace or operating-system audit coverage.
If automatic behavior is absent, ask the user to open /hooks and trust the current FreshRead hook definition. Do not edit files under PLUGIN_DATA manually; use freshread clear for recovery.
Security
- Security scanPassed, 90 of 100Pattern checks on the skill's files.
- Model reviewPassed
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingplugins/freshread/skills/freshread/SKILL.md
Popularity
- Repository stars
- 99
- 7-day trend score
- 23
Stars and forks are for the whole GitHub repository, not this skill alone. Not refreshed yet.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/glaz-j/codex-freshread/freshread)