@orygn/opa-mcp 0.1.1 → 0.1.3

package/CHANGELOG.md

@@ -17,6 +17,53 @@ not part of the public surface and may change in minor releases.
 
 ## [Unreleased]
 
+## [0.1.3] - 2026-05-18
+
+### Added
+
+- **`rego_coverage_gaps` tool.** Runs `opa test --coverage` and returns a
+  per-file breakdown of uncovered line ranges, sorted by coverage ascending
+  so the worst-covered files appear first. Accepts an optional `threshold`
+  to limit output to files below a target coverage percentage. Surfaces
+  `testsPassed`, `testsFailed`, `testsSkipped`, and `overallCoverage` in
+  the envelope alongside the gap report.
+
+- **`rego_security_audit` tool.** Runs regal lint restricted to the
+  `security` and `bugs` categories across one or more policy directories.
+  Returns findings grouped by severity (high / medium) with per-finding
+  remediation guidance. Designed for fleet-wide periodic sweeps rather
+  than per-file style review. Requires regal.
+
+- **`mcp_server_info` tool.** Returns the server name, version, resolved
+  `opa` and `regal` versions, transport type, and Node.js version in a
+  single call. Useful for verifying which server instance an agent is
+  talking to and confirming binary paths resolved correctly.
+
+- **Claude Code install section** in README with the `claude mcp add --env`
+  command. Standing-instructions template (`examples/CLAUDE.md`) and
+  PostToolUse hook config (`examples/claude-code-hook.json`) for policy
+  repos using Claude Code.
+
+- **Node 24 added to CI matrix.** Unit tests now run on Node 20, 22, and 24
+  across Ubuntu, macOS, and Windows.
+
+### Changed
+
+- All em dashes in source comments replaced with `--` (U+002D pairs).
+  No behavior change; cosmetic consistency fix.
+
+## [0.1.2] - 2026-05-18
+
+### Fixed
+
+- `rego_capabilities` with `current: true` or a `version` argument now
+  returns only builtin names, a count, future keywords, and features by
+  default (`names_only: true`). Previously the full spec payload -- type
+  signatures, documentation, and metadata for every builtin -- routinely
+  exceeded the 100 KB `maxResponseBytes` cap and returned a useless
+  `__truncated` envelope. Pass `names_only: false` to retrieve the
+  complete payload when type signatures or documentation are needed.
+
 ## [0.1.1] - 2026-05-09
 
 ### Fixed
@@ -145,6 +192,8 @@ wrappers end-to-end. CI matrix: Ubuntu, macOS, and Windows on Node
 20 and 22, plus CodeQL security scanning and weekly Dependabot updates
 for npm, GitHub Actions, and Docker base images.
 
-[Unreleased]: https://github.com/OrygnsCode/opa-mcp-server/compare/v0.1.1...HEAD
+[Unreleased]: https://github.com/OrygnsCode/opa-mcp-server/compare/v0.1.3...HEAD
+[0.1.3]: https://github.com/OrygnsCode/opa-mcp-server/compare/v0.1.2...v0.1.3
+[0.1.2]: https://github.com/OrygnsCode/opa-mcp-server/releases/tag/v0.1.2
 [0.1.1]: https://github.com/OrygnsCode/opa-mcp-server/releases/tag/v0.1.1
 [0.1.0]: https://github.com/OrygnsCode/opa-mcp-server/releases/tag/v0.1.0

package/README.md

@@ -17,10 +17,10 @@ environment.
 +--------------------+  MCP / stdio  +-----------------+  spawn / HTTP  +------------------+
 |  Claude · Cursor · | ------------> |  @orygn/opa-mcp | -------------> |  opa · regal ·   |
 |   VS Code · ...    | <------------ |                 | <------------- |  OPA REST API    |
-+--------------------+   32 tools    +-----------------+                +------------------+
++--------------------+   35 tools    +-----------------+                +------------------+
 ```
 
-> **Status:** v0.1.1. Tool surface, error codes, and
+> **Status:** v0.1.3. Tool surface, error codes, and
 > environment variables follow [SemVer](https://semver.org/) from
 > v0.1.0 forward.
 
@@ -100,20 +100,10 @@ recognize the gap this fills.
 The server runs locally over stdio. Pick the install path that matches
 your client.
 
-### Claude Desktop / Claude Code
+### Claude Desktop
 
-The fastest path is the Smithery one-liner:
-
-```bash
-npx -y @smithery/cli install @orygn/opa-mcp --client claude
-```
-
-Or download `opa-mcp.mcpb` from the
-[latest release](https://github.com/OrygnsCode/opa-mcp-server/releases/latest)
-and double-click it.
-
-If you prefer to edit `claude_desktop_config.json` by hand, the snippet
-lives in [`examples/claude-desktop.json`](./examples/claude-desktop.json):
+Edit `claude_desktop_config.json` directly (or copy from
+[`examples/claude-desktop.json`](./examples/claude-desktop.json)):
 
 ```json
 {
@@ -136,6 +126,47 @@ lives in [`examples/claude-desktop.json`](./examples/claude-desktop.json):
 > [first-time install gotcha](#-first-time-install-gotcha-read-this-if-you-used-npx-or-the-global-install)
 > below. Windows users substitute `C:\\path\\to\\opa.exe`.
 
+Or download `opa-mcp.mcpb` from the
+[latest release](https://github.com/OrygnsCode/opa-mcp-server/releases/latest)
+and double-click it.
+
+Alternatively, use the Smithery one-liner:
+
+```bash
+npx -y @smithery/cli install @orygn/opa-mcp --client claude
+```
+
+### Claude Code (CLI)
+
+Register the server for the current project with `claude mcp add`:
+
+```bash
+claude mcp add \
+  --env OPA_BINARY=/usr/local/bin/opa \
+  --env REGAL_BINARY=/usr/local/bin/regal \
+  --env OPA_MCP_ALLOWED_PATHS=/path/to/your/policies \
+  opa -- npx -y @orygn/opa-mcp
+```
+
+This writes the config into `.mcp.json` at your project root and is
+picked up automatically on every `claude` session in that directory.
+Add `--scope user` to register it globally instead.
+
+> Replace the paths with your real absolute paths (same caveat as
+> Claude Desktop above). On Windows use `C:\path\to\opa.exe` syntax.
+
+**Persistent context and auto-checks for policy repos.**
+If you work in an OPA policy repo regularly, two extra files remove
+repetitive setup from every session:
+
+- [`examples/CLAUDE.md`](./examples/CLAUDE.md) -- copy to your repo
+  root or `.claude/CLAUDE.md`. Claude Code loads it every session,
+  so the agent always knows which tools to use and what conventions apply.
+- [`examples/claude-code-hook.json`](./examples/claude-code-hook.json) --
+  merge the `hooks` block into `.claude/settings.json`. Runs `opa check`
+  automatically after any `.rego` file is written, so syntax errors
+  surface immediately without a manual tool call.
+
 ### Cursor
 
 Drop [`examples/cursor.json`](./examples/cursor.json) into either
@@ -369,6 +400,14 @@ the tasks agents are actually asked to do.
 | `rego_generate_test_skeleton` | Given a policy, generate a `_test.rego` skeleton covering each rule.                                                                                  |
 | `rego_describe_policy`        | Summarize what a policy does, its inputs, decisions, and assumptions.                                                                                 |
 | `rego_suggest_fix`            | For a failed `rego_check` or `rego_lint`, propose minimal patches.                                                                                    |
+| `rego_coverage_gaps`          | Run `opa test --coverage` and return per-file uncovered line ranges, sorted worst first. Use `threshold` to focus on files below a target percentage. |
+| `rego_security_audit`         | Run regal lint restricted to `security` and `bugs` categories across a directory. Returns severity-grouped findings with remediation guidance.        |
+
+### Category F: Meta
+
+| Tool              | What it does                                                                                                                                                                         |
+| ----------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ |
+| `mcp_server_info` | Return server name, version, resolved `opa`/`regal` versions, transport type, and Node.js version in one call. Useful for verifying which server instance the agent is connected to. |
 
 ## Prompts
 
@@ -452,7 +491,8 @@ Agent: Done. Policy `rbac` is live on staging at $OPA_URL.
 │                          ├── tools/evaluation/        ─┤                              │
 │                          ├── tools/bundles/           ─┼─── lib/opa-cli.ts ──┐        │
 │                          ├── tools/server-management/ ─┤                     │        │
-│                          ├── tools/helpers/           ─┘                     │        │
+│                          ├── tools/helpers/           ─┤                     │        │
+│                          ├── tools/meta/              ─┘                     │        │
 │                          │                                                   ▼        │
 │                          │                              lib/subprocess.ts ──┴── opa   │
 │                          │                              lib/regal-cli.ts   ───── regal│
@@ -554,11 +594,10 @@ upgrade Regal: `brew upgrade regal` or download from the
 
 **`directory-package-mismatch` violation when linting inline source.**
 
-When you pass `source` rather than `paths` to `rego_lint`, Regal sees a
-randomized temp-file path that can't possibly match your declared package
-path. The diagnostic is an artifact of inline linting, not a real issue.
-Disable the rule for inline workflows or lint via `paths` against the
-real on-disk file when you want canonical signal.
+Since v0.1.1, the server auto-disables this rule for inline-source calls.
+If you see it, you are running an older version -- upgrade to v0.1.1 or
+later. To get canonical signal on this rule, lint via `paths` against the
+real on-disk file instead of passing `source` directly.
 
 **Where are the logs?**
 
@@ -588,8 +627,8 @@ npm run build             # compile to dist/
 ```
 
 CI runs lint, typecheck, build, and unit tests on every push and PR
-across Ubuntu, macOS, and Windows on Node 20 and 22. Integration tests
-run on Linux against pinned `opa` and `regal` releases.
+across Ubuntu, macOS, and Windows on Node 20, 22, and 24. Integration
+tests run on Linux against pinned `opa` and `regal` releases.
 
 For the full contributor workflow (adding tools, naming conventions,
 logging discipline, release process), see [CONTRIBUTING.md](./CONTRIBUTING.md).

package/dist/config.d.ts

@@ -15,7 +15,7 @@ declare const ConfigSchema: z.ZodObject<{
     /**
      * Allow-listed root directories for file path inputs. Tools that accept
      * filesystem paths reject anything outside these roots. Empty by
-     * default — file-based tools refuse to read from disk until the
+     * default -- file-based tools refuse to read from disk until the
      * operator explicitly opts in via `OPA_MCP_ALLOWED_PATHS`.
      */
     allowedPaths: z.ZodDefault<z.ZodArray<z.ZodString, "many">>;

package/dist/config.js

@@ -1,7 +1,7 @@
 /**
  * Centralized configuration loaded from environment variables.
  *
- * Environment variables are the only configuration surface — there is no
+ * Environment variables are the only configuration surface -- there is no
  * config file, no flags. This matches how MCP clients (Claude Desktop,
  * Cursor, VS Code) pass config via the `env` object in their JSON.
  */
@@ -24,7 +24,7 @@ const ConfigSchema = z.object({
     /**
      * Allow-listed root directories for file path inputs. Tools that accept
      * filesystem paths reject anything outside these roots. Empty by
-     * default — file-based tools refuse to read from disk until the
+     * default -- file-based tools refuse to read from disk until the
      * operator explicitly opts in via `OPA_MCP_ALLOWED_PATHS`.
      */
     allowedPaths: z.array(z.string()).default([]),

package/dist/constants.d.ts

@@ -0,0 +1,8 @@
+/**
+ * Server identity constants shared between the entry point and the
+ * mcp_server_info tool. Defined here to avoid a circular import
+ * (server.ts -> tools -> meta -> server.ts).
+ */
+export declare const SERVER_NAME = "orygn-opa-mcp";
+export declare const SERVER_VERSION = "0.1.3";
+//# sourceMappingURL=constants.d.ts.map

package/dist/constants.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"constants.d.ts","sourceRoot":"","sources":["../src/constants.ts"],"names":[],"mappings":"AAAA;;;;GAIG;AACH,eAAO,MAAM,WAAW,kBAAkB,CAAC;AAC3C,eAAO,MAAM,cAAc,UAAU,CAAC"}

package/dist/constants.js

@@ -0,0 +1,8 @@
+/**
+ * Server identity constants shared between the entry point and the
+ * mcp_server_info tool. Defined here to avoid a circular import
+ * (server.ts -> tools -> meta -> server.ts).
+ */
+export const SERVER_NAME = 'orygn-opa-mcp';
+export const SERVER_VERSION = '0.1.3';
+//# sourceMappingURL=constants.js.map

package/dist/constants.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"constants.js","sourceRoot":"","sources":["../src/constants.ts"],"names":[],"mappings":"AAAA;;;;GAIG;AACH,MAAM,CAAC,MAAM,WAAW,GAAG,eAAe,CAAC;AAC3C,MAAM,CAAC,MAAM,cAAc,GAAG,OAAO,CAAC"}

package/dist/lib/errors.d.ts

@@ -1,7 +1,7 @@
 /**
  * Helpers for building structured tool errors.
  *
- * Tools should never throw raw exceptions to the MCP layer — every failure
+ * Tools should never throw raw exceptions to the MCP layer -- every failure
  * mode returns a `ToolEnvelope` with `ok: false` and a structured `error`.
  */
 import type { ToolEnvelope, ToolErrorCode } from '../types.js';

package/dist/lib/logger.js

@@ -1,7 +1,7 @@
 /**
  * File-only logger.
  *
- * Stdout is reserved for MCP protocol traffic — writing anywhere else
+ * Stdout is reserved for MCP protocol traffic -- writing anywhere else
  * would corrupt the JSON-RPC stream. This logger appends to a file with
  * level filtering, and never touches stdout.
  *
@@ -37,7 +37,7 @@ function write(level, msg, ctx) {
     }
     catch {
         // Logging must never throw. If the log file is unreachable, drop the
-        // line silently — better than crashing the server.
+        // line silently -- better than crashing the server.
     }
 }
 export const logger = {

package/dist/lib/logger.js.map

@@ -1 +1 @@
-{"version":3,"file":"logger.js","sourceRoot":"","sources":["../../src/lib/logger.ts"],"names":[],"mappings":"AAAA;;;;;;;;GAQG;AACH,OAAO,EAAE,cAAc,EAAE,MAAM,SAAS,CAAC;AAIzC,MAAM,cAAc,GAA0B;IAC5C,KAAK,EAAE,EAAE;IACT,IAAI,EAAE,EAAE;IACR,IAAI,EAAE,EAAE;IACR,KAAK,EAAE,EAAE;CACV,CAAC;AAOF,kDAAkD;AAClD,IAAI,KAA8B,CAAC;AAEnC,MAAM,UAAU,UAAU,CAAC,IAAY,EAAE,KAAY;IACnD,KAAK,GAAG,EAAE,IAAI,EAAE,QAAQ,EAAE,KAAK,EAAE,CAAC;AACpC,CAAC;AAED,SAAS,KAAK,CAAC,KAAY,EAAE,GAAW,EAAE,GAA6B;IACrE,0EAA0E;IAC1E,yDAAyD;IACzD,IAAI,CAAC,KAAK;QAAE,OAAO;IACnB,IAAI,cAAc,CAAC,KAAK,CAAC,GAAG,cAAc,CAAC,KAAK,CAAC,QAAQ,CAAC;QAAE,OAAO;IAEnE,MAAM,KAAK,GAAG;QACZ,EAAE,EAAE,IAAI,IAAI,EAAE,CAAC,WAAW,EAAE;QAC5B,KAAK;QACL,GAAG;QACH,GAAG,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,GAAG,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;KACxB,CAAC;IACF,IAAI,CAAC;QACH,cAAc,CAAC,KAAK,CAAC,IAAI,EAAE,GAAG,IAAI,CAAC,SAAS,CAAC,KAAK,CAAC,IAAI,EAAE,MAAM,CAAC,CAAC;IACnE,CAAC;IAAC,MAAM,CAAC;QACP,qEAAqE;QACrE,mDAAmD;IACrD,CAAC;AACH,CAAC;AAED,MAAM,CAAC,MAAM,MAAM,GAAG;IACpB,KAAK,EAAE,CAAC,GAAW,EAAE,GAA6B,EAAQ,EAAE;QAC1D,KAAK,CAAC,OAAO,EAAE,GAAG,EAAE,GAAG,CAAC,CAAC;IAC3B,CAAC;IACD,IAAI,EAAE,CAAC,GAAW,EAAE,GAA6B,EAAQ,EAAE;QACzD,KAAK,CAAC,MAAM,EAAE,GAAG,EAAE,GAAG,CAAC,CAAC;IAC1B,CAAC;IACD,IAAI,EAAE,CAAC,GAAW,EAAE,GAA6B,EAAQ,EAAE;QACzD,KAAK,CAAC,MAAM,EAAE,GAAG,EAAE,GAAG,CAAC,CAAC;IAC1B,CAAC;IACD,KAAK,EAAE,CAAC,GAAW,EAAE,GAA6B,EAAQ,EAAE;QAC1D,KAAK,CAAC,OAAO,EAAE,GAAG,EAAE,GAAG,CAAC,CAAC;IAC3B,CAAC;CACF,CAAC"}
+{"version":3,"file":"logger.js","sourceRoot":"","sources":["../../src/lib/logger.ts"],"names":[],"mappings":"AAAA;;;;;;;;GAQG;AACH,OAAO,EAAE,cAAc,EAAE,MAAM,SAAS,CAAC;AAIzC,MAAM,cAAc,GAA0B;IAC5C,KAAK,EAAE,EAAE;IACT,IAAI,EAAE,EAAE;IACR,IAAI,EAAE,EAAE;IACR,KAAK,EAAE,EAAE;CACV,CAAC;AAOF,kDAAkD;AAClD,IAAI,KAA8B,CAAC;AAEnC,MAAM,UAAU,UAAU,CAAC,IAAY,EAAE,KAAY;IACnD,KAAK,GAAG,EAAE,IAAI,EAAE,QAAQ,EAAE,KAAK,EAAE,CAAC;AACpC,CAAC;AAED,SAAS,KAAK,CAAC,KAAY,EAAE,GAAW,EAAE,GAA6B;IACrE,0EAA0E;IAC1E,yDAAyD;IACzD,IAAI,CAAC,KAAK;QAAE,OAAO;IACnB,IAAI,cAAc,CAAC,KAAK,CAAC,GAAG,cAAc,CAAC,KAAK,CAAC,QAAQ,CAAC;QAAE,OAAO;IAEnE,MAAM,KAAK,GAAG;QACZ,EAAE,EAAE,IAAI,IAAI,EAAE,CAAC,WAAW,EAAE;QAC5B,KAAK;QACL,GAAG;QACH,GAAG,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,GAAG,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;KACxB,CAAC;IACF,IAAI,CAAC;QACH,cAAc,CAAC,KAAK,CAAC,IAAI,EAAE,GAAG,IAAI,CAAC,SAAS,CAAC,KAAK,CAAC,IAAI,EAAE,MAAM,CAAC,CAAC;IACnE,CAAC;IAAC,MAAM,CAAC;QACP,qEAAqE;QACrE,oDAAoD;IACtD,CAAC;AACH,CAAC;AAED,MAAM,CAAC,MAAM,MAAM,GAAG;IACpB,KAAK,EAAE,CAAC,GAAW,EAAE,GAA6B,EAAQ,EAAE;QAC1D,KAAK,CAAC,OAAO,EAAE,GAAG,EAAE,GAAG,CAAC,CAAC;IAC3B,CAAC;IACD,IAAI,EAAE,CAAC,GAAW,EAAE,GAA6B,EAAQ,EAAE;QACzD,KAAK,CAAC,MAAM,EAAE,GAAG,EAAE,GAAG,CAAC,CAAC;IAC1B,CAAC;IACD,IAAI,EAAE,CAAC,GAAW,EAAE,GAA6B,EAAQ,EAAE;QACzD,KAAK,CAAC,MAAM,EAAE,GAAG,EAAE,GAAG,CAAC,CAAC;IAC1B,CAAC;IACD,KAAK,EAAE,CAAC,GAAW,EAAE,GAA6B,EAAQ,EAAE;QAC1D,KAAK,CAAC,OAAO,EAAE,GAAG,EAAE,GAAG,CAAC,CAAC;IAC3B,CAAC;CACF,CAAC"}

package/dist/lib/opa-cli.d.ts

@@ -138,7 +138,7 @@ export interface SignInput {
 /**
  * Wrapper around the local `opa` binary.
  *
- * Methods do not throw on `opa` errors — non-zero exit codes are
+ * Methods do not throw on `opa` errors -- non-zero exit codes are
  * surfaced on the returned `SpawnResult` so tools can map them to
  * structured error envelopes at their layer. They DO throw on
  * caller-side bugs (e.g. an empty `paths` array passed where one is
@@ -154,7 +154,7 @@ export declare class OpaCli {
     version(): Promise<string | null>;
     /**
      * Format Rego source. Stdout contains the formatted output. Reports
-     * `exitCode: 0` even when no changes are needed — callers compare
+     * `exitCode: 0` even when no changes are needed -- callers compare
      * input vs output to detect a no-op.
      */
     fmt(input: FmtInput): Promise<SpawnResult>;
@@ -162,7 +162,7 @@ export declare class OpaCli {
      * Type-check Rego. Returns `exitCode: 0` and empty `stdout` when the
      * policy is valid. On failure, `exitCode` is non-zero and the JSON
      * error report is written to **stderr** (this is OPA's actual
-     * behavior — tools must read `stderr`, not `stdout`, for `check`
+     * behavior -- tools must read `stderr`, not `stdout`, for `check`
      * diagnostics). Either inline `source` or one or more `paths` must be
      * provided.
      */

package/dist/lib/opa-cli.js

@@ -10,7 +10,7 @@
  * The `-` (read-from-stdin) convention some opa subcommands document is
  * unreliable on Windows, where `-` is treated as a literal filename.
  *
- * The methods do not parse stdout into typed result objects — different
+ * The methods do not parse stdout into typed result objects -- different
  * tools need different shapes of the same output and a one-size parser
  * would force conversions both ways. Stdout is JSON whenever
  * `--format=json` is set, and tools call `JSON.parse` on it themselves.
@@ -23,7 +23,7 @@ import { runBinary } from './subprocess.js';
 /**
  * Wrapper around the local `opa` binary.
  *
- * Methods do not throw on `opa` errors — non-zero exit codes are
+ * Methods do not throw on `opa` errors -- non-zero exit codes are
  * surfaced on the returned `SpawnResult` so tools can map them to
  * structured error envelopes at their layer. They DO throw on
  * caller-side bugs (e.g. an empty `paths` array passed where one is
@@ -48,7 +48,7 @@ export class OpaCli {
     // ─── Authoring ───────────────────────────────────────────────────────
     /**
      * Format Rego source. Stdout contains the formatted output. Reports
-     * `exitCode: 0` even when no changes are needed — callers compare
+     * `exitCode: 0` even when no changes are needed -- callers compare
      * input vs output to detect a no-op.
      */
     async fmt(input) {
@@ -58,7 +58,7 @@ export class OpaCli {
      * Type-check Rego. Returns `exitCode: 0` and empty `stdout` when the
      * policy is valid. On failure, `exitCode` is non-zero and the JSON
      * error report is written to **stderr** (this is OPA's actual
-     * behavior — tools must read `stderr`, not `stdout`, for `check`
+     * behavior -- tools must read `stderr`, not `stdout`, for `check`
      * diagnostics). Either inline `source` or one or more `paths` must be
      * provided.
      */

package/dist/lib/opa-client.d.ts

@@ -29,7 +29,7 @@ export interface RequestOptions {
     body?: unknown;
     /**
      * Raw string body sent verbatim. Used for endpoints that accept
-     * non-JSON content — notably `PUT /v1/policies/{id}` which expects
+     * non-JSON content -- notably `PUT /v1/policies/{id}` which expects
      * Rego source as `text/plain`.
      */
     rawBody?: string;

package/dist/lib/regal-cli.d.ts

@@ -39,7 +39,7 @@ export interface LintInput {
 /**
  * Wrapper around the local `regal` binary.
  *
- * Like `OpaCli`, methods do not throw on Regal-side errors — the exit
+ * Like `OpaCli`, methods do not throw on Regal-side errors -- the exit
  * code on the returned `SpawnResult` is the signal. Inline source is
  * always written to a temp file because `regal lint` does not read
  * from stdin.
@@ -58,9 +58,9 @@ export declare class RegalCli {
      * provided.
      *
      * Regal's exit codes:
-     * - 0 — no findings at or above `failLevel`
-     * - 3 — findings present
-     * - non-zero other — Regal-internal failure (config error, etc.)
+     * - 0 -- no findings at or above `failLevel`
+     * - 3 -- findings present
+     * - non-zero other -- Regal-internal failure (config error, etc.)
      *
      * When called with inline `source`, location-bound rules whose
      * verdict depends on the on-disk path (currently

package/dist/lib/regal-cli.js

@@ -1,7 +1,7 @@
 /**
  * Wrapper around the optional `regal` binary (Rego linter, by Styra).
  *
- * Regal is OPTIONAL — only the `rego_lint` tool requires it. Other
+ * Regal is OPTIONAL -- only the `rego_lint` tool requires it. Other
  * tools work without Regal installed. If absent, `rego_lint` returns a
  * structured `REGAL_NOT_FOUND` error with an install hint.
  */
@@ -21,7 +21,7 @@ export const INLINE_SOURCE_FALSE_POSITIVE_RULES = ['directory-package-mismatch']
 /**
  * Wrapper around the local `regal` binary.
  *
- * Like `OpaCli`, methods do not throw on Regal-side errors — the exit
+ * Like `OpaCli`, methods do not throw on Regal-side errors -- the exit
  * code on the returned `SpawnResult` is the signal. Inline source is
  * always written to a temp file because `regal lint` does not read
  * from stdin.
@@ -48,9 +48,9 @@ export class RegalCli {
      * provided.
      *
      * Regal's exit codes:
-     * - 0 — no findings at or above `failLevel`
-     * - 3 — findings present
-     * - non-zero other — Regal-internal failure (config error, etc.)
+     * - 0 -- no findings at or above `failLevel`
+     * - 3 -- findings present
+     * - non-zero other -- Regal-internal failure (config error, etc.)
      *
      * When called with inline `source`, location-bound rules whose
      * verdict depends on the on-disk path (currently

package/dist/lib/subprocess.d.ts

@@ -19,7 +19,7 @@ export interface SpawnResult {
     durationMs: number;
 }
 /**
- * Run a binary and return its captured output. Never throws — failures
+ * Run a binary and return its captured output. Never throws -- failures
  * are reflected in `exitCode` / `timedOut`.
  */
 export declare function runBinary(binary: string, opts: SpawnOptions): Promise<SpawnResult>;

package/dist/lib/subprocess.js

@@ -1,14 +1,14 @@
 /**
  * Subprocess wrapper used by `opa-cli.ts` and `regal-cli.ts`.
  *
- * - Uses argv arrays only (never shell strings) — prevents injection.
+ * - Uses argv arrays only (never shell strings) -- prevents injection.
  * - Hard timeout per invocation, defaulting to config.subprocessTimeoutMs.
  * - Captures stdout / stderr / exit code separately.
  * - Optional stdin payload for piping source code.
  */
 import { spawn } from 'node:child_process';
 /**
- * Run a binary and return its captured output. Never throws — failures
+ * Run a binary and return its captured output. Never throws -- failures
  * are reflected in `exitCode` / `timedOut`.
  */
 export async function runBinary(binary, opts) {

package/dist/lib/tool-helpers.d.ts

@@ -1,7 +1,7 @@
 /**
  * Helpers shared across tool implementations.
  *
- * Each tool follows the same shape — validate input, call into a CLI
+ * Each tool follows the same shape -- validate input, call into a CLI
  * wrapper or HTTP client, map errors to a structured envelope, return
  * the formatted MCP result. The pieces here factor out the parts every
  * tool repeats so each tool file can stay focused on its own logic.

package/dist/prompts/index.js

@@ -1,5 +1,5 @@
 /**
- * MCP Prompts — slash-command-like workflow templates.
+ * MCP Prompts -- slash-command-like workflow templates.
  *
  * Each prompt is a stateless instruction set the agent receives when
  * the user invokes it. They orient the agent toward a specific
@@ -64,7 +64,7 @@ Return a concise review with: pass/fail per item, recommended diffs.`;
 const decisionDebuggingWorkflowPrompt = (args) => {
     return `You are debugging an unexpected Rego decision.
 
-Query: ${args.query ?? '<not provided — ask the user>'}
+Query: ${args.query ?? '<not provided -- ask the user>'}
 User's expectation: ${args.expectation ?? '<ask the user what they expected>'}
 
 Workflow:
@@ -77,13 +77,13 @@ Workflow:
 3. Re-run with \`rego_explain_decision\` to get a structured trace.
    Identify which rules were evaluated and which fired.
 4. The cause is one of:
-   a. **Input mismatch** — the policy expected a different input shape.
+   a. **Input mismatch** -- the policy expected a different input shape.
       Use \`rego_describe_policy\` to list the refs the policy reads
       and confirm each is present in the input.
-   b. **Rule logic** — a guard fired or didn't fire when it should
+   b. **Rule logic** -- a guard fired or didn't fire when it should
       have. Read the trace and explain which rule's body evaluated to
       true/false and why.
-   c. **Default decision** — no rule produced a value, so the default
+   c. **Default decision** -- no rule produced a value, so the default
       kicked in.
 5. Propose the smallest fix: either an input correction or a policy
    change. If a policy change, run \`rego_check\` and \`rego_test\` on the
@@ -115,7 +115,7 @@ export function registerPrompts(server, _config) {
             source: z
                 .string()
                 .optional()
-                .describe('Rego source to review. Optional — agent can ask for it.'),
+                .describe('Rego source to review. Optional -- agent can ask for it.'),
         },
     }, (args) => ({
         messages: [

package/dist/prompts/index.js.map

@@ -1 +1 @@
-{"version":3,"file":"index.js","sourceRoot":"","sources":["../../src/prompts/index.ts"],"names":[],"mappings":"AAAA;;;;;;;GAOG;AACH,OAAO,EAAE,CAAC,EAAE,MAAM,KAAK,CAAC;AAMxB,MAAM,8BAA8B,GAAG,CAAC,IAGvC,EAAU,EAAE;IACX,MAAM,WAAW,GAAG,IAAI,CAAC,WAAW,IAAI,gBAAgB,CAAC;IACzD,MAAM,WAAW,GAAG,IAAI,CAAC,YAAY,IAAI,gCAAgC,CAAC;IAC1E,OAAO;;;;;EAKP,WAAW;;qBAEQ,WAAW;;;;;;;;;;;;;;;gDAegB,CAAC;AACjD,CAAC,CAAC;AAEF,MAAM,2BAA2B,GAAG,CAAC,IAAyB,EAAU,EAAE;IACxE,OAAO;;;EAGP,IAAI,CAAC,MAAM,IAAI,8CAA8C;;;;;;;;;;;;;;;;;;;;;qEAqBM,CAAC;AACtE,CAAC,CAAC;AAEF,MAAM,+BAA+B,GAAG,CAAC,IAGxC,EAAU,EAAE;IACX,OAAO;;SAEA,IAAI,CAAC,KAAK,IAAI,+BAA+B;sBAChC,IAAI,CAAC,WAAW,IAAI,mCAAmC;;;;;;;;;;;;;;;;;;;;;;;;;2CAyBlC,CAAC;AAC5C,CAAC,CAAC;AAEF,MAAM,UAAU,eAAe,CAAC,MAAiB,EAAE,OAAe;IAChE,MAAM,CAAC,cAAc,CACnB,4BAA4B,EAC5B;QACE,KAAK,EAAE,4BAA4B;QACnC,WAAW,EACT,uHAAuH;QACzH,UAAU,EAAE;YACV,WAAW,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,EAAE,CAAC,QAAQ,CAAC,mCAAmC,CAAC;YAChF,YAAY,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,EAAE,CAAC,QAAQ,CAAC,wCAAwC,CAAC;SACvF;KACF,EACD,CAAC,IAAI,EAAE,EAAE,CAAC,CAAC;QACT,QAAQ,EAAE;YACR;gBACE,IAAI,EAAE,MAAM;gBACZ,OAAO,EAAE,EAAE,IAAI,EAAE,MAAM,EAAE,IAAI,EAAE,8BAA8B,CAAC,IAAI,CAAC,EAAE;aACtE;SACF;KACF,CAAC,CACH,CAAC;IAEF,MAAM,CAAC,cAAc,CACnB,yBAAyB,EACzB;QACE,KAAK,EAAE,yBAAyB;QAChC,WAAW,EACT,wHAAwH;QAC1H,UAAU,EAAE;YACV,MAAM,EAAE,CAAC;iBACN,MAAM,EAAE;iBACR,QAAQ,EAAE;iBACV,QAAQ,CAAC,yDAAyD,CAAC;SACvE;KACF,EACD,CAAC,IAAI,EAAE,EAAE,CAAC,CAAC;QACT,QAAQ,EAAE;YACR;gBACE,IAAI,EAAE,MAAM;gBACZ,OAAO,EAAE,EAAE,IAAI,EAAE,MAAM,EAAE,IAAI,EAAE,2BAA2B,CAAC,IAAI,CAAC,EAAE;aACnE;SACF;KACF,CAAC,CACH,CAAC;IAEF,MAAM,CAAC,cAAc,CACnB,6BAA6B,EAC7B;QACE,KAAK,EAAE,6BAA6B;QACpC,WAAW,EACT,2IAA2I;QAC7I,UAAU,EAAE;YACV,KAAK,EAAE,CAAC;iBACL,MAAM,EAAE;iBACR,QAAQ,EAAE;iBACV,QAAQ,CAAC,qDAAqD,CAAC;YAClE,WAAW,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,EAAE,CAAC,QAAQ,CAAC,mCAAmC,CAAC;SACjF;KACF,EACD,CAAC,IAAI,EAAE,EAAE,CAAC,CAAC;QACT,QAAQ,EAAE;YACR;gBACE,IAAI,EAAE,MAAM;gBACZ,OAAO,EAAE,EAAE,IAAI,EAAE,MAAM,EAAE,IAAI,EAAE,+BAA+B,CAAC,IAAI,CAAC,EAAE;aACvE;SACF;KACF,CAAC,CACH,CAAC;AACJ,CAAC"}
+{"version":3,"file":"index.js","sourceRoot":"","sources":["../../src/prompts/index.ts"],"names":[],"mappings":"AAAA;;;;;;;GAOG;AACH,OAAO,EAAE,CAAC,EAAE,MAAM,KAAK,CAAC;AAMxB,MAAM,8BAA8B,GAAG,CAAC,IAGvC,EAAU,EAAE;IACX,MAAM,WAAW,GAAG,IAAI,CAAC,WAAW,IAAI,gBAAgB,CAAC;IACzD,MAAM,WAAW,GAAG,IAAI,CAAC,YAAY,IAAI,gCAAgC,CAAC;IAC1E,OAAO;;;;;EAKP,WAAW;;qBAEQ,WAAW;;;;;;;;;;;;;;;gDAegB,CAAC;AACjD,CAAC,CAAC;AAEF,MAAM,2BAA2B,GAAG,CAAC,IAAyB,EAAU,EAAE;IACxE,OAAO;;;EAGP,IAAI,CAAC,MAAM,IAAI,8CAA8C;;;;;;;;;;;;;;;;;;;;;qEAqBM,CAAC;AACtE,CAAC,CAAC;AAEF,MAAM,+BAA+B,GAAG,CAAC,IAGxC,EAAU,EAAE;IACX,OAAO;;SAEA,IAAI,CAAC,KAAK,IAAI,gCAAgC;sBACjC,IAAI,CAAC,WAAW,IAAI,mCAAmC;;;;;;;;;;;;;;;;;;;;;;;;;2CAyBlC,CAAC;AAC5C,CAAC,CAAC;AAEF,MAAM,UAAU,eAAe,CAAC,MAAiB,EAAE,OAAe;IAChE,MAAM,CAAC,cAAc,CACnB,4BAA4B,EAC5B;QACE,KAAK,EAAE,4BAA4B;QACnC,WAAW,EACT,uHAAuH;QACzH,UAAU,EAAE;YACV,WAAW,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,EAAE,CAAC,QAAQ,CAAC,mCAAmC,CAAC;YAChF,YAAY,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,EAAE,CAAC,QAAQ,CAAC,wCAAwC,CAAC;SACvF;KACF,EACD,CAAC,IAAI,EAAE,EAAE,CAAC,CAAC;QACT,QAAQ,EAAE;YACR;gBACE,IAAI,EAAE,MAAM;gBACZ,OAAO,EAAE,EAAE,IAAI,EAAE,MAAM,EAAE,IAAI,EAAE,8BAA8B,CAAC,IAAI,CAAC,EAAE;aACtE;SACF;KACF,CAAC,CACH,CAAC;IAEF,MAAM,CAAC,cAAc,CACnB,yBAAyB,EACzB;QACE,KAAK,EAAE,yBAAyB;QAChC,WAAW,EACT,wHAAwH;QAC1H,UAAU,EAAE;YACV,MAAM,EAAE,CAAC;iBACN,MAAM,EAAE;iBACR,QAAQ,EAAE;iBACV,QAAQ,CAAC,0DAA0D,CAAC;SACxE;KACF,EACD,CAAC,IAAI,EAAE,EAAE,CAAC,CAAC;QACT,QAAQ,EAAE;YACR;gBACE,IAAI,EAAE,MAAM;gBACZ,OAAO,EAAE,EAAE,IAAI,EAAE,MAAM,EAAE,IAAI,EAAE,2BAA2B,CAAC,IAAI,CAAC,EAAE;aACnE;SACF;KACF,CAAC,CACH,CAAC;IAEF,MAAM,CAAC,cAAc,CACnB,6BAA6B,EAC7B;QACE,KAAK,EAAE,6BAA6B;QACpC,WAAW,EACT,2IAA2I;QAC7I,UAAU,EAAE;YACV,KAAK,EAAE,CAAC;iBACL,MAAM,EAAE;iBACR,QAAQ,EAAE;iBACV,QAAQ,CAAC,qDAAqD,CAAC;YAClE,WAAW,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,EAAE,CAAC,QAAQ,CAAC,mCAAmC,CAAC;SACjF;KACF,EACD,CAAC,IAAI,EAAE,EAAE,CAAC,CAAC;QACT,QAAQ,EAAE;YACR;gBACE,IAAI,EAAE,MAAM;gBACZ,OAAO,EAAE,EAAE,IAAI,EAAE,MAAM,EAAE,IAAI,EAAE,+BAA+B,CAAC,IAAI,CAAC,EAAE;aACvE;SACF;KACF,CAAC,CACH,CAAC;AACJ,CAAC"}

package/dist/resources/index.d.ts

@@ -1,9 +1,9 @@
 /**
- * MCP Resources — read-only references the agent can fetch by URI.
+ * MCP Resources -- read-only references the agent can fetch by URI.
  *
- * - opa://builtins      — derived at read-time from `opa capabilities`.
- * - opa://style-guide   — curated Rego style guide content.
- * - opa://patterns      — pattern library: RBAC, ABAC, K8s admission,
+ * - opa://builtins      -- derived at read-time from `opa capabilities`.
+ * - opa://style-guide   -- curated Rego style guide content.
+ * - opa://patterns      -- pattern library: RBAC, ABAC, K8s admission,
  *                         IaC gates, API authz, rate limiting.
  */
 import type { McpServer } from '@modelcontextprotocol/sdk/server/mcp.js';

package/dist/resources/patterns.d.ts

@@ -5,5 +5,5 @@
  *   - A test
  *   - Common pitfalls
  */
-export declare const PATTERNS = "# Rego pattern library\n\nCommon Rego patterns with working examples, tests, and pitfalls. Each\npattern is self-contained \u2014 copy, adapt, and ship.\n\n---\n\n## 1. Role-based access control (RBAC)\n\n**When to use:** the simplest authorization model. Every user has one\nor more roles; each role grants a set of actions on a set of\nresources. Sufficient for ~80% of internal applications.\n\n```rego\npackage rbac\n\nimport rego.v1\n\ndefault allow := false\n\n# Permissions table \u2014 extend as roles evolve.\npermissions := {\n    \"admin\": {\"read\", \"write\", \"delete\", \"manage_users\"},\n    \"editor\": {\"read\", \"write\"},\n    \"viewer\": {\"read\"},\n}\n\nallow if {\n    some role in input.user.roles\n    input.action in permissions[role]\n}\n\n# Why was it denied? \u2014 useful for audit logs.\ndeny_reasons contains reason if {\n    not allow\n    input.user\n    reason := sprintf(\n        \"user %q has roles %v, none grant %q\",\n        [input.user.id, input.user.roles, input.action],\n    )\n}\n\ndeny_reasons contains \"anonymous request\" if {\n    not allow\n    not input.user\n}\n```\n\n**Test:**\n\n```rego\npackage rbac_test\n\nimport rego.v1\nimport data.rbac\n\ntest_admin_can_delete if {\n    rbac.allow with input as {\n        \"user\": {\"id\": \"alice\", \"roles\": [\"admin\"]},\n        \"action\": \"delete\",\n    }\n}\n\ntest_viewer_cannot_delete if {\n    not rbac.allow with input as {\n        \"user\": {\"id\": \"bob\", \"roles\": [\"viewer\"]},\n        \"action\": \"delete\",\n    }\n}\n\ntest_anonymous_denied if {\n    not rbac.allow with input as {\"action\": \"read\"}\n    \"anonymous request\" in (rbac.deny_reasons with input as {\"action\": \"read\"})\n}\n```\n\n**Pitfalls:**\n- The `permissions` table grows unbounded. Move to data files\n  (`data.permissions`) once you have more than ~20 roles.\n- Roles overlap with groups in real auth systems; map at the boundary\n  rather than carrying both.\n\n---\n\n## 2. Attribute-based access control (ABAC)\n\n**When to use:** when \"who\" alone isn't enough \u2014 decisions also depend\non resource attributes (ownership, tenant, sensitivity) and context\n(time of day, source IP).\n\n```rego\npackage abac\n\nimport rego.v1\n\ndefault allow := false\n\n# A user can read any resource they own.\nallow if {\n    input.action == \"read\"\n    input.resource.owner_id == input.user.id\n}\n\n# A user can read shared resources at their organization.\nallow if {\n    input.action == \"read\"\n    input.resource.shared\n    input.resource.org_id == input.user.org_id\n}\n\n# Admins can do anything within their organization.\nallow if {\n    \"admin\" in input.user.roles\n    input.resource.org_id == input.user.org_id\n}\n\n# Don't show \"secret\" resources to anyone outside the owner's\n# organization, even admins.\nallow := false if {\n    input.resource.classification == \"secret\"\n    input.resource.org_id != input.user.org_id\n}\n```\n\n**Test:**\n\n```rego\npackage abac_test\n\nimport rego.v1\nimport data.abac\n\ntest_owner_reads_own if {\n    abac.allow with input as {\n        \"action\": \"read\",\n        \"user\": {\"id\": \"u1\", \"org_id\": \"o1\"},\n        \"resource\": {\"owner_id\": \"u1\", \"org_id\": \"o1\"},\n    }\n}\n\ntest_admin_blocked_from_secret_in_other_org if {\n    not abac.allow with input as {\n        \"action\": \"read\",\n        \"user\": {\"id\": \"u1\", \"org_id\": \"o1\", \"roles\": [\"admin\"]},\n        \"resource\": {\n            \"owner_id\": \"u2\",\n            \"org_id\": \"o2\",\n            \"classification\": \"secret\",\n        },\n    }\n}\n```\n\n**Pitfalls:**\n- Multiple `allow` rules combine with logical OR. Use explicit\n  `allow := false if ...` to *override* an allow.\n- Don't compute attributes inside the policy; compute them at the\n  boundary and pass via `input`.\n\n---\n\n## 3. Kubernetes admission control\n\n**When to use:** validate or mutate Kubernetes resources at admission\ntime. Run as a Gatekeeper, OPA-as-a-webhook, or Kyverno-equivalent\npolicy layer.\n\n```rego\npackage k8s.admission\n\nimport rego.v1\n\n# Reject pods without resource limits.\ndeny contains msg if {\n    input.request.kind.kind == \"Pod\"\n    container := input.request.object.spec.containers[_]\n    not container.resources.limits.memory\n    msg := sprintf(\n        \"Pod %q container %q is missing resources.limits.memory\",\n        [input.request.object.metadata.name, container.name],\n    )\n}\n\n# Reject privileged containers in production.\ndeny contains msg if {\n    input.request.kind.kind == \"Pod\"\n    input.request.namespace != \"kube-system\"\n    container := input.request.object.spec.containers[_]\n    container.securityContext.privileged == true\n    msg := sprintf(\n        \"privileged containers are not allowed: %q in %q\",\n        [container.name, input.request.object.metadata.name],\n    )\n}\n```\n\n**Pitfalls:**\n- Use `input.request.object` for the resource being admitted; the\n  envelope shape comes from Kubernetes, not your control.\n- For mutations, return a JSON Patch via the `patch` field. Test\n  patches with the actual admission webhook in dry-run mode before\n  enforcing.\n- Iteration order is undefined; never rely on `containers[0]` to mean\n  anything specific.\n\n---\n\n## 4. Infrastructure-as-Code gates (Terraform)\n\n**When to use:** validate Terraform plans before apply. Catch overly\npermissive IAM, public S3 buckets, missing encryption.\n\n```rego\npackage terraform\n\nimport rego.v1\n\n# Reject S3 buckets without server-side encryption.\ndeny contains msg if {\n    resource := input.resource_changes[_]\n    resource.type == \"aws_s3_bucket\"\n    after := resource.change.after\n    not after.server_side_encryption_configuration\n    msg := sprintf(\n        \"S3 bucket %q has no server-side encryption configured\",\n        [resource.address],\n    )\n}\n\n# Reject IAM policies with action \"*\" on resource \"*\".\ndeny contains msg if {\n    resource := input.resource_changes[_]\n    resource.type == \"aws_iam_policy\"\n    policy := json.unmarshal(resource.change.after.policy)\n    statement := policy.Statement[_]\n    statement.Effect == \"Allow\"\n    \"*\" in statement.Action\n    statement.Resource == \"*\"\n    msg := sprintf(\"IAM policy %q grants Allow * on *\", [resource.address])\n}\n```\n\n**Pitfalls:**\n- Terraform plan JSON is verbose and version-specific. Pin the\n  `terraform plan -json` schema you target.\n- `resource_changes[_].change.after` may be `null` for destroys \u2014\n  guard against it.\n- For wide-radius changes, use `opa exec --decision` against a plan\n  file in CI, not the live API.\n\n---\n\n## 5. API authorization (HTTP request gating)\n\n**When to use:** at the API gateway / reverse proxy layer, validate\neach request against the caller's identity and the requested\nendpoint.\n\n```rego\npackage api.authz\n\nimport rego.v1\n\ndefault allow := false\n\n# Public endpoints \u2014 no auth required.\npublic_endpoints := {\n    {\"method\": \"GET\", \"path\": [\"health\"]},\n    {\"method\": \"GET\", \"path\": [\"version\"]},\n}\n\nallow if some _ in public_endpoints; matches_endpoint(_)\n\n# Authenticated reads on resources the user has access to.\nallow if {\n    input.method == \"GET\"\n    input.user\n    user_can_read(input.user, input.path)\n}\n\n# Authenticated writes only with specific scopes.\nallow if {\n    input.method in {\"POST\", \"PUT\", \"PATCH\", \"DELETE\"}\n    input.user\n    \"write\" in input.user.scopes\n    user_can_write(input.user, input.path)\n}\n\nmatches_endpoint(spec) if {\n    spec.method == input.method\n    spec.path == input.path\n}\n\nuser_can_read(user, path) if {\n    path[0] == \"users\"\n    user.id == path[1]\n}\n\nuser_can_write(user, path) if {\n    path[0] == \"users\"\n    user.id == path[1]\n}\n```\n\n**Pitfalls:**\n- `input.path` is typically an array (`[\"users\", \"alice\"]`), not a\n  string. Build it consistently at the gateway.\n- Path-prefix matching is easy; full pattern matching is not. For\n  parameterized routes, decode at the gateway and pass structured\n  fields.\n\n---\n\n## 6. Rate limiting (with sliding window data)\n\n**When to use:** allow N requests per principal per window. Lightweight\nlimit enforcement; for high throughput, push to a dedicated rate\nlimiter.\n\n```rego\npackage rate\n\nimport rego.v1\n\n# Configuration: 100 requests per principal per 60-second window.\nlimit := 100\nwindow_seconds := 60\n\n# input.now is a unix timestamp in nanoseconds.\n# data.requests[principal] is an array of nanosecond timestamps.\n\ncurrent_window := requests if {\n    requests := [t | some t in data.requests[input.principal]; t > input.now - window_seconds * 1000000000]\n}\n\ncount := count(current_window)\n\nallow if count < limit\n\ndeny_reason := sprintf(\n    \"rate limit exceeded: %d requests in last %d seconds (limit %d)\",\n    [count, window_seconds, limit],\n) if not allow\n```\n\n**Pitfalls:**\n- `data.requests` grows unbounded unless the writer prunes outside the\n  window. Schedule prune on every write.\n- This pattern is *advisory* \u2014 under load, two concurrent decisions\n  can both see `count == limit - 1` and both allow. For strict\n  limits, use a Lua/Redis token bucket at the gateway and have OPA\n  validate the token, not count requests.\n\n---\n\n## Where these patterns came from\n\nEach is distilled from production policy code. The full Rego files,\ntests, and policy data fixtures live in this server's GitHub\nrepository under `tests/fixtures/policies/`.\n\nFor more patterns, see:\n\n- OPA Playground: https://play.openpolicyagent.org/\n- Awesome OPA: https://github.com/anderseknert/awesome-opa\n- Styra DAS pattern library: https://docs.styra.com/das/policies\n";
+export declare const PATTERNS = "# Rego pattern library\n\nCommon Rego patterns with working examples, tests, and pitfalls. Each\npattern is self-contained -- copy, adapt, and ship.\n\n---\n\n## 1. Role-based access control (RBAC)\n\n**When to use:** the simplest authorization model. Every user has one\nor more roles; each role grants a set of actions on a set of\nresources. Sufficient for ~80% of internal applications.\n\n```rego\npackage rbac\n\nimport rego.v1\n\ndefault allow := false\n\n# Permissions table -- extend as roles evolve.\npermissions := {\n    \"admin\": {\"read\", \"write\", \"delete\", \"manage_users\"},\n    \"editor\": {\"read\", \"write\"},\n    \"viewer\": {\"read\"},\n}\n\nallow if {\n    some role in input.user.roles\n    input.action in permissions[role]\n}\n\n# Why was it denied? -- useful for audit logs.\ndeny_reasons contains reason if {\n    not allow\n    input.user\n    reason := sprintf(\n        \"user %q has roles %v, none grant %q\",\n        [input.user.id, input.user.roles, input.action],\n    )\n}\n\ndeny_reasons contains \"anonymous request\" if {\n    not allow\n    not input.user\n}\n```\n\n**Test:**\n\n```rego\npackage rbac_test\n\nimport rego.v1\nimport data.rbac\n\ntest_admin_can_delete if {\n    rbac.allow with input as {\n        \"user\": {\"id\": \"alice\", \"roles\": [\"admin\"]},\n        \"action\": \"delete\",\n    }\n}\n\ntest_viewer_cannot_delete if {\n    not rbac.allow with input as {\n        \"user\": {\"id\": \"bob\", \"roles\": [\"viewer\"]},\n        \"action\": \"delete\",\n    }\n}\n\ntest_anonymous_denied if {\n    not rbac.allow with input as {\"action\": \"read\"}\n    \"anonymous request\" in (rbac.deny_reasons with input as {\"action\": \"read\"})\n}\n```\n\n**Pitfalls:**\n- The `permissions` table grows unbounded. Move to data files\n  (`data.permissions`) once you have more than ~20 roles.\n- Roles overlap with groups in real auth systems; map at the boundary\n  rather than carrying both.\n\n---\n\n## 2. Attribute-based access control (ABAC)\n\n**When to use:** when \"who\" alone isn't enough -- decisions also depend\non resource attributes (ownership, tenant, sensitivity) and context\n(time of day, source IP).\n\n```rego\npackage abac\n\nimport rego.v1\n\ndefault allow := false\n\n# A user can read any resource they own.\nallow if {\n    input.action == \"read\"\n    input.resource.owner_id == input.user.id\n}\n\n# A user can read shared resources at their organization.\nallow if {\n    input.action == \"read\"\n    input.resource.shared\n    input.resource.org_id == input.user.org_id\n}\n\n# Admins can do anything within their organization.\nallow if {\n    \"admin\" in input.user.roles\n    input.resource.org_id == input.user.org_id\n}\n\n# Don't show \"secret\" resources to anyone outside the owner's\n# organization, even admins.\nallow := false if {\n    input.resource.classification == \"secret\"\n    input.resource.org_id != input.user.org_id\n}\n```\n\n**Test:**\n\n```rego\npackage abac_test\n\nimport rego.v1\nimport data.abac\n\ntest_owner_reads_own if {\n    abac.allow with input as {\n        \"action\": \"read\",\n        \"user\": {\"id\": \"u1\", \"org_id\": \"o1\"},\n        \"resource\": {\"owner_id\": \"u1\", \"org_id\": \"o1\"},\n    }\n}\n\ntest_admin_blocked_from_secret_in_other_org if {\n    not abac.allow with input as {\n        \"action\": \"read\",\n        \"user\": {\"id\": \"u1\", \"org_id\": \"o1\", \"roles\": [\"admin\"]},\n        \"resource\": {\n            \"owner_id\": \"u2\",\n            \"org_id\": \"o2\",\n            \"classification\": \"secret\",\n        },\n    }\n}\n```\n\n**Pitfalls:**\n- Multiple `allow` rules combine with logical OR. Use explicit\n  `allow := false if ...` to *override* an allow.\n- Don't compute attributes inside the policy; compute them at the\n  boundary and pass via `input`.\n\n---\n\n## 3. Kubernetes admission control\n\n**When to use:** validate or mutate Kubernetes resources at admission\ntime. Run as a Gatekeeper, OPA-as-a-webhook, or Kyverno-equivalent\npolicy layer.\n\n```rego\npackage k8s.admission\n\nimport rego.v1\n\n# Reject pods without resource limits.\ndeny contains msg if {\n    input.request.kind.kind == \"Pod\"\n    container := input.request.object.spec.containers[_]\n    not container.resources.limits.memory\n    msg := sprintf(\n        \"Pod %q container %q is missing resources.limits.memory\",\n        [input.request.object.metadata.name, container.name],\n    )\n}\n\n# Reject privileged containers in production.\ndeny contains msg if {\n    input.request.kind.kind == \"Pod\"\n    input.request.namespace != \"kube-system\"\n    container := input.request.object.spec.containers[_]\n    container.securityContext.privileged == true\n    msg := sprintf(\n        \"privileged containers are not allowed: %q in %q\",\n        [container.name, input.request.object.metadata.name],\n    )\n}\n```\n\n**Pitfalls:**\n- Use `input.request.object` for the resource being admitted; the\n  envelope shape comes from Kubernetes, not your control.\n- For mutations, return a JSON Patch via the `patch` field. Test\n  patches with the actual admission webhook in dry-run mode before\n  enforcing.\n- Iteration order is undefined; never rely on `containers[0]` to mean\n  anything specific.\n\n---\n\n## 4. Infrastructure-as-Code gates (Terraform)\n\n**When to use:** validate Terraform plans before apply. Catch overly\npermissive IAM, public S3 buckets, missing encryption.\n\n```rego\npackage terraform\n\nimport rego.v1\n\n# Reject S3 buckets without server-side encryption.\ndeny contains msg if {\n    resource := input.resource_changes[_]\n    resource.type == \"aws_s3_bucket\"\n    after := resource.change.after\n    not after.server_side_encryption_configuration\n    msg := sprintf(\n        \"S3 bucket %q has no server-side encryption configured\",\n        [resource.address],\n    )\n}\n\n# Reject IAM policies with action \"*\" on resource \"*\".\ndeny contains msg if {\n    resource := input.resource_changes[_]\n    resource.type == \"aws_iam_policy\"\n    policy := json.unmarshal(resource.change.after.policy)\n    statement := policy.Statement[_]\n    statement.Effect == \"Allow\"\n    \"*\" in statement.Action\n    statement.Resource == \"*\"\n    msg := sprintf(\"IAM policy %q grants Allow * on *\", [resource.address])\n}\n```\n\n**Pitfalls:**\n- Terraform plan JSON is verbose and version-specific. Pin the\n  `terraform plan -json` schema you target.\n- `resource_changes[_].change.after` may be `null` for destroys --\n  guard against it.\n- For wide-radius changes, use `opa exec --decision` against a plan\n  file in CI, not the live API.\n\n---\n\n## 5. API authorization (HTTP request gating)\n\n**When to use:** at the API gateway / reverse proxy layer, validate\neach request against the caller's identity and the requested\nendpoint.\n\n```rego\npackage api.authz\n\nimport rego.v1\n\ndefault allow := false\n\n# Public endpoints -- no auth required.\npublic_endpoints := {\n    {\"method\": \"GET\", \"path\": [\"health\"]},\n    {\"method\": \"GET\", \"path\": [\"version\"]},\n}\n\nallow if some _ in public_endpoints; matches_endpoint(_)\n\n# Authenticated reads on resources the user has access to.\nallow if {\n    input.method == \"GET\"\n    input.user\n    user_can_read(input.user, input.path)\n}\n\n# Authenticated writes only with specific scopes.\nallow if {\n    input.method in {\"POST\", \"PUT\", \"PATCH\", \"DELETE\"}\n    input.user\n    \"write\" in input.user.scopes\n    user_can_write(input.user, input.path)\n}\n\nmatches_endpoint(spec) if {\n    spec.method == input.method\n    spec.path == input.path\n}\n\nuser_can_read(user, path) if {\n    path[0] == \"users\"\n    user.id == path[1]\n}\n\nuser_can_write(user, path) if {\n    path[0] == \"users\"\n    user.id == path[1]\n}\n```\n\n**Pitfalls:**\n- `input.path` is typically an array (`[\"users\", \"alice\"]`), not a\n  string. Build it consistently at the gateway.\n- Path-prefix matching is easy; full pattern matching is not. For\n  parameterized routes, decode at the gateway and pass structured\n  fields.\n\n---\n\n## 6. Rate limiting (with sliding window data)\n\n**When to use:** allow N requests per principal per window. Lightweight\nlimit enforcement; for high throughput, push to a dedicated rate\nlimiter.\n\n```rego\npackage rate\n\nimport rego.v1\n\n# Configuration: 100 requests per principal per 60-second window.\nlimit := 100\nwindow_seconds := 60\n\n# input.now is a unix timestamp in nanoseconds.\n# data.requests[principal] is an array of nanosecond timestamps.\n\ncurrent_window := requests if {\n    requests := [t | some t in data.requests[input.principal]; t > input.now - window_seconds * 1000000000]\n}\n\ncount := count(current_window)\n\nallow if count < limit\n\ndeny_reason := sprintf(\n    \"rate limit exceeded: %d requests in last %d seconds (limit %d)\",\n    [count, window_seconds, limit],\n) if not allow\n```\n\n**Pitfalls:**\n- `data.requests` grows unbounded unless the writer prunes outside the\n  window. Schedule prune on every write.\n- This pattern is *advisory* -- under load, two concurrent decisions\n  can both see `count == limit - 1` and both allow. For strict\n  limits, use a Lua/Redis token bucket at the gateway and have OPA\n  validate the token, not count requests.\n\n---\n\n## Where these patterns came from\n\nEach is distilled from production policy code. The full Rego files,\ntests, and policy data fixtures live in this server's GitHub\nrepository under `tests/fixtures/policies/`.\n\nFor more patterns, see:\n\n- OPA Playground: https://play.openpolicyagent.org/\n- Awesome OPA: https://github.com/anderseknert/awesome-opa\n- Styra DAS pattern library: https://docs.styra.com/das/policies\n";
 //# sourceMappingURL=patterns.d.ts.map

package/dist/resources/patterns.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"patterns.d.ts","sourceRoot":"","sources":["../../src/resources/patterns.ts"],"names":[],"mappings":"AAAA;;;;;;GAMG;AACH,eAAO,MAAM,QAAQ,glTA8WpB,CAAC"}
+{"version":3,"file":"patterns.d.ts","sourceRoot":"","sources":["../../src/resources/patterns.ts"],"names":[],"mappings":"AAAA;;;;;;GAMG;AACH,eAAO,MAAM,QAAQ,ojTA8WpB,CAAC"}