By ProfSynapse
Enterprise guardrails for agentic tools: CRUA (create/read/update/archive — never delete), agent workspace with checkout/publish, sync-aware safety for OneDrive/Drive/Dropbox folders, and policy-driven command/content blocking.
Publish a checked-out working copy back over the original (archives prior version)
Restore a file from the guardrails archive (latest or a chosen version)
Show open agent-workspace checkouts and recent guardrail activity
Audit report — what the guardrails allowed, asked about, and blocked recently
CRUA workflow for editing Office and proprietary documents (docx, xlsx, pptx, pdf) safely. Use when asked to modify, update, fill in, or rewrite a document file — checkout converts it to markdown/csv in _workspace/, you edit the copy, publish archives the old version and replaces the original. Never edit proprietary formats in place.
Handling Google Docs/Sheets/Slides files that appear in a Google Drive for desktop synced folder as .gdoc/.gsheet/.gslides files. Use when a task involves one of these extensions or when the user asks to edit a Google Doc from their local Drive folder.
Recovering lost or overwritten files from the guardrails archive store. Use when the user says a file was deleted, overwritten, truncated, or "messed up", or asks to roll back to an earlier version or undo a recent file operation.
Safety rules for working inside cloud-synced folders — OneDrive, SharePoint synced libraries, Google Drive for desktop, Dropbox. Use whenever a task touches files under a synced folder, before bulk operations, or when a file read returns empty/garbage content for a file that clearly should have data.
Executes bash commands
Hook triggers when Bash tool is used
Modifies files
Hook triggers on file write and edit operations
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
Make agentic tools (Claude Code, Cowork) safe on a real computer, including the OneDrive/SharePoint/Google Drive/Dropbox folders synced to it, with one plugin install.
The core promise: nothing is ever destroyed.
rm and every destructive equivalent is blocked and redirected to
agw archive: a reversible, versioned move into an archive store.>, mv, cp, or tee would clobber, so the
promise holds even when the agent bypasses the Write tool._workspace/, and agw publish
archives the old version before replacing the original, with conflict
detection if a human edited it in the meantime..gdoc pointer stubs (the classic synced-
folder data-loss traps) are detected and protected.agw restore or agw undo./plugin marketplace add https://github.com/ProfSynapse/agentic-guardrails-plugin.git
/plugin install agentic-guardrails@synaptic-guardrails
If Claude's marketplace UI rejects ProfSynapse/agentic-guardrails-plugin, use
the full GitHub URL above instead of the owner/repo shorthand.
Requires Python 3.9+ as python3. Optional: pandoc (docx↔markdown) and
openpyxl (xlsx→csv) for high-fidelity document checkout; without them files
are checked out in plain-copy mode. Fleet rollout: see
enterprise/DEPLOYMENT.md.
| Piece | Purpose |
|---|---|
hooks/ | PreToolUse/PostToolUse/SessionStart wiring, the enforcement surface (works in Claude Code and Cowork) |
scripts/claude/ | Thin Claude adapter: tool call → neutral ToolEvent, decision → hook JSON. Fails closed (any internal error → "ask", never silent allow) |
scripts/core/ | Platform-neutral policy engine: shell parser (substitutions, bash -c, xargs, wrappers, decode-pipes), folder profiles, archive store, audit log with secret redaction |
scripts/agw/ + bin/agw | The agw CLI ("agent workspace"): scan, checkout, diff, publish, archive, restore, undo, move, snapshot, status, log, doctor, plus office for targeted in-place docx/xlsx/pptx edits (replace-text, set-cell, append-rows) with automatic pre-image snapshots |
policies/ | Editable YAML rules: command rules, content/snippet rules (regex → deny/ask), path zones. Per-machine drop-ins in ~/.agw/policies.d/ |
skills/ | Teach the agent the workflows: agent-workspace, synced-folders, gdocs-bridge, restore |
commands/ | /agw-status, /agw-publish, /agw-restore, /guardrails-report |
enterprise/ | Managed-settings template + deployment guide |
Denied primitives always come with a safe replacement in the denial message, so the agent self-corrects instead of fighting the rails:
| Instead of | The agent uses |
|---|---|
rm file | agw archive file (reversible) |
editing report.docx in place | agw checkout → edit markdown → agw publish |
python -c openpyxl one-liners | agw office set-cell / replace-text / append-rows |
mv (untracked) | agw move (logged, undoable) |
| bulk folder surgery | agw snapshot first, then work |
Exception: rm of purely regenerable build/dependency dirs (node_modules,
dist, .venv, __pycache__...) is allowed at standard and above (pointless
and huge to archive). strict archives even those. The list is extensible via
settings.regenerable_globs.
Escalations (ask): git checkout -- <file>, shrink-suspicious writes
(replacing a large file with tiny content), reading cloud-only placeholders,
publish conflicts, agw prune/apply/hydrate, reading credential-type
files (.env, keys, ~/.aws...), files whose content prescan finds secrets or
"CONFIDENTIAL" markings ("this might contain a password, confirm"), and
recursive credential-keyword searches. Combining a credential file with a
network tool in one command (curl -d @.env ...) is denied as exfiltration. Hard denies: rm/shred/
find -delete, git push --force / reset --hard / clean -f, dd to
devices, mkfs, sudo, decode-to-shell and download-to-shell pipes,
destructive SQL/interpreter one-liners, writes to .gdoc stubs, placeholders,
protected zones, the plugin itself, and the archive store.
npx claudepluginhub profsynapse/agentic-guardrails-plugin --plugin agentic-guardrailsProfessor Synapse — a router that summons expert agents on demand, with persistent agent-tagged memory and a summon-gate that keeps work on-persona. Your own agents and memory live in the plugin's data dir and survive updates.
Harness-native ECC plugin for engineering teams - 67 agents, 271 skills, 92 legacy command shims, reusable hooks, rules, MCP conventions, and operator workflows for Claude Code plus adjacent agent harnesses
Core skills library for Claude Code: TDD, debugging, collaboration patterns, and proven techniques
Tools to maintain and improve CLAUDE.md files - audit quality, capture session learnings, and keep project memory current.
v9.44.1 — Patch release for Gemini environment/version detection and qwen auth gating. Run /octo:setup.
Superpowers Plus core skills library for Claude Code: planning, execution routing, TDD, debugging, and collaboration workflows
Persistent file-based planning for AI coding agents. Crash-proof markdown plans (task_plan.md, findings.md, progress.md) that survive context loss and /clear, with an opt-in completion gate and multi-agent shared state. Manus-style. Works with Claude Code, Codex CLI, Cursor, Kiro, OpenCode and 60+ agents via the SKILL.md standard. Includes Arabic, German, Spanish, and Chinese (Simplified and Traditional).