@gotgenes/pi-permission-system

Permission enforcement extension for the Pi coding agent that provides centralized, deterministic permission gates over tool, bash, MCP, skill, and special operations.

Fork notice: This package is a full fork of MasuRii/pi-permission-system, published to npm as @gotgenes/pi-permission-system. It has diverged substantially from upstream in config format, internal architecture, and permission model.

What It Does

• Hides disallowed tools before the agent starts — no wasted turns probing for blocked tools
• Enforces allow / ask / deny at tool-call time with UI confirmation dialogs
• Controls bash commands with wildcard pattern matching (git *: ask, rm -rf *: deny)
• Gates MCP and skill access at server, tool, and skill-name granularity
• Protects sensitive file patterns — cross-cutting path rules deny .env, ~/.ssh/*, etc. across all tools and bash at once, matching both the path as referenced and its symlink-resolved form so a deny cannot be evaded through a symlink alias
• Guards external paths — prompts before file tools or bash commands reach outside cwd
• Fails closed — an internal gate error blocks the tool (with a gate_error review-log entry and a matching permissions:decision broadcast), and an unparseable bash command — or an indirection wrapper that hides the gated command (bash -c/eval, sudo, env, xargs, find -exec, …) — prompts (ask) rather than passing silently
• Forwards prompts from subagents — ask policies work even in non-UI execution contexts
• Broadcasts UI prompt events — permissions:ui_prompt fires only when the permission system is about to invoke the active user-facing permission UI, and every prompt it announces — including one forwarded up from a subagent — is answered by a permissions:decision on the same bus
• Native @gotgenes/pi-subagents integration — in-process child sessions register with the permission system automatically, enabling per-agent policy enforcement and ask-state forwarding to the parent UI without configuration

Install

pi install npm:@gotgenes/pi-permission-system

Quick Start

1. Create the global config file at ~/.pi/agent/extensions/pi-permission-system/config.json:

   {
     "permission": {
       "*": "allow",
       "path": {
         "*": "allow",
         "*.env": "deny",
         "*.env.*": "deny",
         "*.env.example": "allow"
       },
       "bash": {
         "*": "ask",
         "rm -rf *": "deny",
         "sudo *": "ask"
       },
       "external_directory": "ask"
     }
   }
   

2. Start Pi — the extension automatically loads and enforces your policy.

All permissions use one of three states:

State	Behavior
allow	Permits the action silently
deny	Blocks the action with an error message
ask	Prompts the user for confirmation via UI

When the dialog prompts, you can approve once or approve a pattern for the rest of the session. In an interactive TUI session the prompt is an inline keybind dialog — y approve, s approve for this session, n deny, r deny with a reason — where each hotkey arms and a second press confirms (configurable via doublePressToConfirm). The prompt shows one fact per line — who is asking, the tool, the matched rule, the value being decided — within a row budget, so a large tool input cannot take over the transcript; Ctrl+O (app.tools.expand) expands it to the complete request. See docs/configuration.md for the hotkeys and docs/session-approvals.md for session-scoped rules and pattern suggestions.

The path surface is a cross-cutting gate that applies to all file access — Pi tools, bash commands, MCP calls, and extension tools alike. Extension and MCP tools that operate on paths (via input.path, MCP's input.arguments.path, or a registered access extractor) are gated by default, so a path deny cannot be overridden by a per-tool allow — making it the right place to protect sensitive files like .env or ~/.ssh/* from every tool at once. A path pattern matches both the path as the agent references it and its canonical (symlink-resolved) form, so a deny still fires when a symlink aliases a sensitive target.

For per-tool path patterns (read, write, edit, find, grep, ls), patterns are matched against the file path from input.path. This lets you express rules like "allow reads but deny .env files" at the individual tool level. Like the cross-cutting path surface, per-tool patterns match both the referenced path and its canonical (symlink-resolved) form, so a per-tool deny resists symlink-alias evasion. When Pi's current working directory is known, relative path inputs also match their cwd-normalized absolute form, so src/App.jsx can match both src/* and /workspace/project/*.

The external_directory surface is the CWD-boundary gate: it decides whether reaching outside the working tree is allowed, and accepts a pattern map so you can allow specific outside-CWD directories without opening up all external access. This is the right surface for silencing repeated prompts on a local cache like ~/.cargo/registry — allow it here, not on path:

{
  "permission": {
    "external_directory": {
      "*": "ask",
      "~/.cargo/registry/*": "allow"
    }
  }
}

The trailing * is greedy and crosses subdirectory boundaries, so it allows every file beneath the directory; a bare ~/.cargo/registry matches only the directory entry itself.

Four layers compose with most-restrictive-wins: path (cross-cutting) → external_directory (CWD boundary) → per-tool patterns → bash command patterns. Because ask is more restrictive than allow, a path allow cannot loosen an external_directory: ask boundary — allow outside-CWD directories on external_directory. See docs/configuration.md for the full recipe.

Configuration

Config lives in one JSON file per scope:

Scope	Path
Global	~/.pi/agent/extensions/pi-permission-system/config.json
Project	<cwd>/.pi/extensions/pi-permission-system/config.json

Project overrides global; per-agent YAML frontmatter overrides both. Project config (policy and runtime knobs) is loaded only once the project is trusted — in an untrusted directory only global config applies, so an untrusted repository cannot loosen your global policy (see Upgrading).

Within a surface map like bash or mcp, last matching rule wins — put broad catch-alls first and specific overrides after.

The optional shellTools field records which non-bash tools carry shell semantics (e.g. an exec_command tool that replaces native bash), so they are gated at full parity with native bash — see docs/configuration.md.

The optional authorizerChain field names registered case-by-case decision links (e.g. a light model judge) to consult when a request lands on ask, ahead of the interactive prompt. A downstream extension registers a link via getPermissionsService(sessionId).registerAuthorizer(name, authorize); it decides nothing until you name it here (opt-in), config order fixes the chain order, and the chain owner caps any link's allow on external_directory/path to keep it within your policy — see docs/configuration.md. A subagent's ask is reviewed by the chain of the session serving it, one hop up, rather than inside the subagent — see the same section. @gotgenes/pi-permission-model-judge is a first-party reference implementation of such a link — a deny-first reviewer that auto-denies mistyped out-of-directory paths.

For the full reference — all surfaces, runtime knobs, per-agent overrides, merge semantics, and common recipes — see docs/configuration.md.

Upgrading

22.0.0 — project config requires project trust

Project-scoped configuration (the project config.json and project-agent frontmatter — both permission policy and runtime knobs such as yoloMode) is now loaded only when Pi reports the project as trusted. In an untrusted directory, only global config applies; a skip is surfaced with a warning and a project_trust.skipped review-log entry. Grant project trust (or set defaultProjectTrust) to load a project's config. See docs/migration/0644-project-trust-gating.md.

16.0.0 — the bash gate now fails closed

The permission gate fails closed: an internal gate error blocks the tool (with a gate_error review-log entry) instead of running it ungated, and a non-empty bash command that cannot be parsed resolves to ask (sentinel <unparseable-bash-command>) rather than falling through to a permissive top-level *. Commands that previously slipped through silently on the error or empty-parse path now block or prompt.

If you relied on the old permissive behavior for bash, set an explicit permissive bash policy — "bash": { "*": "allow" } — which also suppresses the new startup warning emitted when a top-level "*": "allow" leaves bash ungated.

Scope and non-goals

Purpose. An agent takes many actions, most of them benign, but some of which need a human to confirm they are safe or correct. This package routes your attention to those, and turns each ruling into deterministic, reusable policy — enforced at the host level rather than by asking the model to police itself.

In scope. Hardening the gates against bypass, fail-closed corrections (breaking ones included), named opt-in extension seams for downstream packages, and structural work backed by a written decision record.

Non-goals.

• Sandboxing. This is a decision layer, not a sandbox — it decides and records, it does not isolate. If a dangerous action is reachable through an allowed tool, policy has to restrict it explicitly.
• Deciding project trust. A policy enforcer, not a trust oracle: whether a project is trusted is Pi's decision and yours, and this package observes it.
• Permissive defaults, trust profiles, or workflow presets. Your risk profile is not knowable from here, so defaults are least-privilege and common policies ship as documented recipes rather than preset keywords.
• Guessing what is sensitive. No built-in secret denylist, and log redaction is key-name-structural rather than predictive — a redactor that silently misses a key invites treating the log as safe to share.
• Model judgment in the core. This package makes no LLM call and holds no model config; model-assisted judging attaches as a chain link over the authorizer seam instead. A link decides nothing until you name it in authorizerChain, and its allow on an excluded surface is downgraded to defer.

The architecture doc carries the full inventory, with the decision record behind each entry.

One decision is still open. How policy may enter the system — which channels are admissible, and with what precedence — is being worked out in issue #639, along with whether a capability model replaces the current surface list. Several requested widenings are parked on it rather than declined, durable persistence of an approval among them.

Where adjacent requests belong. True isolation of a permitted action → an agent sandbox. Model-assisted judging of an ask → a chain link over the authorizer seam; @gotgenes/pi-permission-model-judge is the first-party one, and judges mistyped paths. Approve-and-steer, edit diffs, and risk explanations → a downstream package over the permissions:decision event and the presentation seams.

Documentation

Document	Contents
docs/configuration.md	Full policy reference, runtime knobs, per-agent overrides, recipes
docs/session-approvals.md	Session-scoped rules, pattern suggestions, bash arity table
docs/cross-extension-api.md	Cross-extension service accessor, event bus integration, prompt and decision broadcasts
docs/subagent-integration.md	The subagent adapter convention, permission forwarding, coexistence with subagent extensions
docs/guides/permission-frontmatter-for-subagent-extensions.md	Convention guide for subagent extension authors
docs/opencode-compatibility.md	OpenCode compatibility — shared concepts, divergences, porting guide
docs/troubleshooting.md	Common issues, diagnostic logging, threat model
docs/migration/legacy-to-flat.md	Migration from pre-v2 config layout
docs/migration/strict-config-validation.md	Strict config validation (breaking) — rejected configs, and the cross-scope fail-closed clamp
docs/migration/0644-project-trust-gating.md	Project-trust gating (breaking) — project config loads only after project trust
docs/migration/0745-prompt-payload-contracts.md	Prompt payload contracts (breaking) — the forwarded wire, the ui_prompt broadcast, and the deprecated preview caps
docs/migration/0746-review-log-fields.md	Review-log fields (breaking) — message replaced by request facts, and the reviewLogFieldMaxWidth bound
docs/migration/0794-keyed-service-locator.md	Keyed service locator (breaking) — getPermissionsService(sessionId), and the repeating ready event

Development

pnpm run check       # Type-check TypeScript (no emit)
pnpm run lint        # Biome + ESLint + lint:md
pnpm run lint:md     # rumdl on README and docs
pnpm run test        # Run tests from ./test
pnpm run test:watch  # Run tests in watch mode

Pre-commit hooks

This project uses prek to run Biome, ESLint, and rumdl on staged files before each commit. Run pnpm install to set up hooks automatically.

Acknowledgments

This project began as a fork of MasuRii/pi-permission-system. Thank you to MasuRii for the original work that made this possible.

Thank you to the OpenCode team for the permission model design that inspired the flat config format and evaluation semantics used in this extension.

License

MIT