@optimuslabs/pii-scrub-staging 0.0.0-stage → 0.1.0

package/README.md

@@ -1,3 +1,22 @@
-# Temporary Holding Version
+# @optimuslabs/pii-scrub
 
-This version is a temporary placeholder for this package. An operational version to replace this has been submitted for review and is awaiting a staged release.
+Redacts secrets, and optionally personal data (email names, home-directory usernames), from what an
+Optimus endpoint sends, before the request leaves the machine. Used by `@optimuslabs/runtime`
+(tool calls), `@optimuslabs/harness-map` (config file inventory) and `@optimuslabs/discovery`
+(which sends the version). Design: `docs/pii-scrubbing/DESIGN.md`.
+
+```ts
+import { SCRUBBER_VERSION, scrub, scrubText, scrubValue } from "@optimuslabs/pii-scrub";
+
+scrubValue({ env: { BRAVE_API_KEY: "BSA..." } });      // { env: { BRAVE_API_KEY: "[OL_REDACTED:CREDENTIAL_ENV_VAR]" } }
+scrubText("key sk-ant-api03-...", { fingerprintKey }); // "key sk-ant-api03-[OL_REDACTED:ANTHROPIC_KEY:<24 hex>]"
+scrubText("ada@gmail.com", { personalData: true });    // "[OL_REDACTED:EMAIL]@gmail.com"
+scrub(payload).hits;                                   // { ANTHROPIC_KEY: 1, ... }; placeholders already there are not hits
+```
+
+The rules are in `rules.json`, which the backend reads too; every pattern works in both the
+JavaScript and Python regex engines. Only vendor-format secrets are fingerprinted (HMAC-SHA256
+under the org key, 96 bits), and only when the caller passes the key.
+
+The staging build publishes as `@optimuslabs/pii-scrub-staging`, the production build as
+`@optimuslabs/pii-scrub`.

package/dist/index.d.ts

@@ -0,0 +1,60 @@
+/**
+ * Redact secrets, and optionally personal data, from what an endpoint sends to Optimus.
+ *
+ * Runs on the endpoint, just before a request is built, so the value never leaves the machine.
+ * A value is replaced with a typed placeholder rather than dropped, so the backend still knows a
+ * credential was there and of what kind:
+ *
+ *   sk-ant-api03-[OL_REDACTED:ANTHROPIC_KEY:<fp>]   a vendor key keeps its published prefix
+ *   [OL_REDACTED:CREDENTIAL_ENV_VAR]                a value under a credential-named key
+ *   [OL_REDACTED:ENV_VALUE]                         any other value in an MCP server's env block
+ *   [OL_REDACTED:EMAIL]@gmail.com                   an email inside content keeps its domain
+ *   /Users/[OL_REDACTED:USERNAME]/src               a home directory keeps its path
+ *
+ * `<fp>` is a 96-bit fingerprint (24 hex characters, HMAC-SHA256 under the org's key), on
+ * vendor-format secrets only, so the same key is recognisable across configs and machines
+ * without being held. A credential-named value with no known format could be a guessable
+ * password, so it gets none. Without a key, nothing gets one.
+ *
+ * Structure first, then text: a 29-character API key matches no format, but nobody names a
+ * variable `BRAVE_API_KEY` unless it holds one. A placeholder already in the input is left as it
+ * is and is not a hit, so scrubbing twice changes nothing and counts nothing.
+ */
+export { SECRET_FORMATS } from "./rules.js";
+/** This scrubber's version (the package's), which every sender includes with what it scrubbed. */
+export declare const SCRUBBER_VERSION: string;
+export declare const PLACEHOLDER_PREFIX = "[OL_REDACTED:";
+export interface ScrubOptions {
+    /**
+     * Also redact personal data: the name part of emails and the username in home directories. On
+     * for tool-call content; off for config files, where account emails are identity (agent owners,
+     * tenant approval, the personal-email finding) and stay readable.
+     */
+    personalData?: boolean;
+    /** The org's fingerprint key, hex. Absent until the endpoint has received it: no fingerprints. */
+    fingerprintKey?: string;
+}
+/** How many of each type were redacted. Placeholders already in the input are not counted. */
+export type ScrubHits = Record<string, number>;
+export interface ScrubResult<T> {
+    value: T;
+    hits: ScrubHits;
+}
+export declare function placeholder(type: string, fingerprint?: string): string;
+/**
+ * A deep copy of `value` with every string scrubbed and every value under a credential-named key
+ * redacted, and how many of each type were redacted. Non-string leaves are returned as they are.
+ */
+export declare function scrub<T>(value: T, options?: ScrubOptions): ScrubResult<T>;
+/** `scrub(value).value`. */
+export declare function scrubValue<T>(value: T, options?: ScrubOptions): T;
+/** A copy of `text` with every secret, and with `personalData` emails and usernames, redacted. */
+export declare function scrubText(text: string, options?: ScrubOptions): string;
+/**
+ * A copy of `record` with each of `fields` that is a string scrubbed as text, the rest untouched:
+ * for log lines and run records (`command`, `message`, `detail`), where the fields are free text.
+ */
+export declare function scrubFields<T extends Record<string, unknown>>(record: T, fields: readonly string[], options?: ScrubOptions): T;
+/** Whether `value` holds a placeholder anywhere (to refuse writing scrubbed data back to disk). */
+export declare function containsPlaceholder(value: unknown): boolean;
+//# sourceMappingURL=index.d.ts.map

package/dist/index.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../src/index.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;;;;GAqBG;AAmBH,OAAO,EAAE,cAAc,EAAE,MAAM,YAAY,CAAC;AAE5C,kGAAkG;AAClG,eAAO,MAAM,gBAAgB,EAAE,MAA2F,CAAC;AAE3H,eAAO,MAAM,kBAAkB,kBAAkB,CAAC;AAElD,MAAM,WAAW,YAAY;IAC3B;;;;OAIG;IACH,YAAY,CAAC,EAAE,OAAO,CAAC;IACvB,kGAAkG;IAClG,cAAc,CAAC,EAAE,MAAM,CAAC;CACzB;AAED,8FAA8F;AAC9F,MAAM,MAAM,SAAS,GAAG,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC,CAAC;AAE/C,MAAM,WAAW,WAAW,CAAC,CAAC;IAC5B,KAAK,EAAE,CAAC,CAAC;IACT,IAAI,EAAE,SAAS,CAAC;CACjB;AAWD,wBAAgB,WAAW,CAAC,IAAI,EAAE,MAAM,EAAE,WAAW,CAAC,EAAE,MAAM,GAAG,MAAM,CAEtE;AAgID;;;GAGG;AACH,wBAAgB,KAAK,CAAC,CAAC,EAAE,KAAK,EAAE,CAAC,EAAE,OAAO,GAAE,YAAiB,GAAG,WAAW,CAAC,CAAC,CAAC,CAG7E;AAED,4BAA4B;AAC5B,wBAAgB,UAAU,CAAC,CAAC,EAAE,KAAK,EAAE,CAAC,EAAE,OAAO,GAAE,YAAiB,GAAG,CAAC,CAErE;AAED,kGAAkG;AAClG,wBAAgB,SAAS,CAAC,IAAI,EAAE,MAAM,EAAE,OAAO,GAAE,YAAiB,GAAG,MAAM,CAE1E;AAED;;;GAGG;AACH,wBAAgB,WAAW,CAAC,CAAC,SAAS,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,EAC3D,MAAM,EAAE,CAAC,EACT,MAAM,EAAE,SAAS,MAAM,EAAE,EACzB,OAAO,GAAE,YAAiB,GACzB,CAAC,CAMH;AAED,mGAAmG;AACnG,wBAAgB,mBAAmB,CAAC,KAAK,EAAE,OAAO,GAAG,OAAO,CAO3D"}

package/dist/index.js

@@ -0,0 +1,191 @@
+/**
+ * Redact secrets, and optionally personal data, from what an endpoint sends to Optimus.
+ *
+ * Runs on the endpoint, just before a request is built, so the value never leaves the machine.
+ * A value is replaced with a typed placeholder rather than dropped, so the backend still knows a
+ * credential was there and of what kind:
+ *
+ *   sk-ant-api03-[OL_REDACTED:ANTHROPIC_KEY:<fp>]   a vendor key keeps its published prefix
+ *   [OL_REDACTED:CREDENTIAL_ENV_VAR]                a value under a credential-named key
+ *   [OL_REDACTED:ENV_VALUE]                         any other value in an MCP server's env block
+ *   [OL_REDACTED:EMAIL]@gmail.com                   an email inside content keeps its domain
+ *   /Users/[OL_REDACTED:USERNAME]/src               a home directory keeps its path
+ *
+ * `<fp>` is a 96-bit fingerprint (24 hex characters, HMAC-SHA256 under the org's key), on
+ * vendor-format secrets only, so the same key is recognisable across configs and machines
+ * without being held. A credential-named value with no known format could be a guessable
+ * password, so it gets none. Without a key, nothing gets one.
+ *
+ * Structure first, then text: a 29-character API key matches no format, but nobody names a
+ * variable `BRAVE_API_KEY` unless it holds one. A placeholder already in the input is left as it
+ * is and is not a hit, so scrubbing twice changes nothing and counts nothing.
+ */
+import { createHmac } from "node:crypto";
+import { createRequire } from "node:module";
+import { CONTAINS_REFERENCE, CREDENTIAL_ASSIGNMENT, CREDENTIAL_FLAG, CREDENTIAL_KEY, EMAIL, ENV_REFERENCE, HOME_DIRECTORY, MCP_SERVER_KEYS, MCP_VALUE_BLOCKS, MIN_CREDENTIAL_LENGTH, RESERVED_EMAIL_DOMAIN, SECRET_FORMATS, } from "./rules.js";
+export { SECRET_FORMATS } from "./rules.js";
+/** This scrubber's version (the package's), which every sender includes with what it scrubbed. */
+export const SCRUBBER_VERSION = createRequire(import.meta.url)("../package.json").version;
+export const PLACEHOLDER_PREFIX = "[OL_REDACTED:";
+/** How deep a payload may nest before the scrubber stops descending (and redacts the rest). */
+const MAX_DEPTH = 32;
+const FINGERPRINT_HEX_LENGTH = 24;
+export function placeholder(type, fingerprint) {
+    return fingerprint ? `${PLACEHOLDER_PREFIX}${type}:${fingerprint}]` : `${PLACEHOLDER_PREFIX}${type}]`;
+}
+function redacted(pass, type, fingerprint) {
+    pass.hits[type] = (pass.hits[type] ?? 0) + 1;
+    return placeholder(type, fingerprint);
+}
+function fingerprintOf(pass, secret) {
+    const key = pass.options.fingerprintKey;
+    if (!key)
+        return undefined;
+    return createHmac("sha256", Buffer.from(key, "hex")).update(secret).digest("hex").slice(0, FINGERPRINT_HEX_LENGTH);
+}
+/** A value worth redacting under a credential-named key: not a flag, a reference or a placeholder. */
+function isCredentialValue(value) {
+    const trimmed = value.trim();
+    return (trimmed.length >= MIN_CREDENTIAL_LENGTH &&
+        !/\s/.test(trimmed) &&
+        !ENV_REFERENCE.test(trimmed) &&
+        // A placeholder, or the front of one that a text rule cut at its closing `]`.
+        !trimmed.includes(PLACEHOLDER_PREFIX));
+}
+function redactFormats(text, pass) {
+    let out = text;
+    for (const format of SECRET_FORMATS) {
+        out = out.replace(format.pattern, (match, prefix) => {
+            const fingerprint = format.fingerprint ? fingerprintOf(pass, match) : undefined;
+            return `${typeof prefix === "string" ? prefix : ""}${redacted(pass, format.type, fingerprint)}`;
+        });
+    }
+    return out;
+}
+function redactAssignments(text, pass) {
+    return text.replace(CREDENTIAL_ASSIGNMENT, (match, name, value) => isCredentialValue(value) ? `${name}${redacted(pass, "CREDENTIAL_ENV_VAR")}` : match);
+}
+function redactPersonalData(text, pass) {
+    // `git@github.com:` is an SSH remote, and a reserved domain is documentation: neither is a person.
+    const emailsDone = text.replace(EMAIL, (match, domain) => match.startsWith("git@") || RESERVED_EMAIL_DOMAIN.test(domain)
+        ? match
+        : `${redacted(pass, "EMAIL")}@${domain}`);
+    return emailsDone.replace(HOME_DIRECTORY, (_match, root) => `${root}${redacted(pass, "USERNAME")}`);
+}
+function scrubString(text, pass) {
+    if (!text)
+        return text;
+    const out = redactAssignments(redactFormats(text, pass), pass);
+    return pass.options.personalData ? redactPersonalData(out, pass) : out;
+}
+function scrubLeaf(value, pass, key) {
+    if (key !== undefined && CREDENTIAL_KEY.test(key) && isCredentialValue(value)) {
+        // A recognised format keeps its prefix, type and fingerprint; anything else is just a credential.
+        const formatted = redactFormats(value, pass);
+        return formatted !== value ? formatted : redacted(pass, "CREDENTIAL_ENV_VAR");
+    }
+    return scrubString(value, pass);
+}
+function isPlainObject(value) {
+    return Boolean(value) && typeof value === "object" && !Array.isArray(value);
+}
+/**
+ * A value in an MCP server's env or headers block. Every one is redacted, because a server's
+ * secrets sit there under whatever name its author chose: a known format keeps its prefix, type
+ * and fingerprint; a credential-named one is a credential; anything else is typed by the block
+ * (`ENV_VALUE`, `HEADER_VALUE`), which the backend does not read as a secret. A reference to a
+ * secret stored elsewhere (`${TOKEN}`, `Bearer ${TOKEN}`) stays readable.
+ */
+function redactBlockValue(value, name, ordinaryType, pass) {
+    if (!value.trim() || CONTAINS_REFERENCE.test(value) || ENV_REFERENCE.test(value.trim()))
+        return value;
+    if (value.includes(PLACEHOLDER_PREFIX))
+        return value;
+    const formatted = redactFormats(value, pass);
+    if (formatted !== value)
+        return formatted;
+    return redacted(pass, CREDENTIAL_KEY.test(name) ? "CREDENTIAL_ENV_VAR" : ordinaryType);
+}
+function walkValueBlock(block, ordinaryType, pass, depth) {
+    const out = {};
+    for (const [name, v] of Object.entries(block)) {
+        out[name] = typeof v === "string" ? redactBlockValue(v, name, ordinaryType, pass) : walk(v, pass, name, depth + 1);
+    }
+    return out;
+}
+/** `["--api-key", "BSA..."]`: the argument after a credential flag is the credential. */
+function walkArgs(args, pass, depth) {
+    return args.map((arg, i) => {
+        const flag = typeof args[i - 1] === "string" ? CREDENTIAL_FLAG.exec(args[i - 1]) : null;
+        if (flag && CREDENTIAL_KEY.test(flag[1]) && typeof arg === "string" && isCredentialValue(arg)) {
+            const formatted = redactFormats(arg, pass);
+            return formatted !== arg ? formatted : redacted(pass, "CREDENTIAL_ENV_VAR");
+        }
+        return walk(arg, pass, "args", depth + 1);
+    });
+}
+/** One child of an object: an MCP value block and `args` get their structural rule. */
+function walkChild(key, value, inServer, pass, depth) {
+    const blockType = inServer ? MCP_VALUE_BLOCKS[key] : undefined;
+    if (blockType && isPlainObject(value))
+        return walkValueBlock(value, blockType, pass, depth);
+    if (key === "args" && Array.isArray(value))
+        return walkArgs(value, pass, depth);
+    return walk(value, pass, key, depth + 1);
+}
+function walkObject(obj, pass, depth) {
+    const inServer = Object.keys(obj).some((k) => MCP_SERVER_KEYS.has(k));
+    const out = {};
+    for (const [k, v] of Object.entries(obj))
+        out[k] = walkChild(k, v, inServer, pass, depth);
+    return out;
+}
+function walk(value, pass, key, depth) {
+    if (depth > MAX_DEPTH)
+        return redacted(pass, "TOO_DEEP");
+    if (Array.isArray(value))
+        return value.map((item) => walk(item, pass, key, depth + 1));
+    if (isPlainObject(value))
+        return walkObject(value, pass, depth);
+    return typeof value === "string" ? scrubLeaf(value, pass, key) : value;
+}
+/**
+ * A deep copy of `value` with every string scrubbed and every value under a credential-named key
+ * redacted, and how many of each type were redacted. Non-string leaves are returned as they are.
+ */
+export function scrub(value, options = {}) {
+    const pass = { options, hits: {} };
+    return { value: walk(value, pass, undefined, 0), hits: pass.hits };
+}
+/** `scrub(value).value`. */
+export function scrubValue(value, options = {}) {
+    return scrub(value, options).value;
+}
+/** A copy of `text` with every secret, and with `personalData` emails and usernames, redacted. */
+export function scrubText(text, options = {}) {
+    return scrubString(text, { options, hits: {} });
+}
+/**
+ * A copy of `record` with each of `fields` that is a string scrubbed as text, the rest untouched:
+ * for log lines and run records (`command`, `message`, `detail`), where the fields are free text.
+ */
+export function scrubFields(record, fields, options = {}) {
+    const out = { ...record };
+    for (const field of fields) {
+        if (typeof out[field] === "string")
+            out[field] = scrubText(out[field], options);
+    }
+    return out;
+}
+/** Whether `value` holds a placeholder anywhere (to refuse writing scrubbed data back to disk). */
+export function containsPlaceholder(value) {
+    if (typeof value === "string")
+        return value.includes(PLACEHOLDER_PREFIX);
+    try {
+        return JSON.stringify(value ?? null).includes(PLACEHOLDER_PREFIX);
+    }
+    catch {
+        return false;
+    }
+}
+//# sourceMappingURL=index.js.map

package/dist/index.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"index.js","sourceRoot":"","sources":["../src/index.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;;;;GAqBG;AAEH,OAAO,EAAE,UAAU,EAAE,MAAM,aAAa,CAAC;AACzC,OAAO,EAAE,aAAa,EAAE,MAAM,aAAa,CAAC;AAC5C,OAAO,EACL,kBAAkB,EAClB,qBAAqB,EACrB,eAAe,EACf,cAAc,EACd,KAAK,EACL,aAAa,EACb,cAAc,EACd,eAAe,EACf,gBAAgB,EAChB,qBAAqB,EACrB,qBAAqB,EACrB,cAAc,GACf,MAAM,YAAY,CAAC;AAEpB,OAAO,EAAE,cAAc,EAAE,MAAM,YAAY,CAAC;AAE5C,kGAAkG;AAClG,MAAM,CAAC,MAAM,gBAAgB,GAAY,aAAa,CAAC,MAAM,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC,iBAAiB,CAAyB,CAAC,OAAO,CAAC;AAE3H,MAAM,CAAC,MAAM,kBAAkB,GAAG,eAAe,CAAC;AA0BlD,+FAA+F;AAC/F,MAAM,SAAS,GAAG,EAAE,CAAC;AACrB,MAAM,sBAAsB,GAAG,EAAE,CAAC;AAElC,MAAM,UAAU,WAAW,CAAC,IAAY,EAAE,WAAoB;IAC5D,OAAO,WAAW,CAAC,CAAC,CAAC,GAAG,kBAAkB,GAAG,IAAI,IAAI,WAAW,GAAG,CAAC,CAAC,CAAC,GAAG,kBAAkB,GAAG,IAAI,GAAG,CAAC;AACxG,CAAC;AAED,SAAS,QAAQ,CAAC,IAAU,EAAE,IAAY,EAAE,WAAoB;IAC9D,IAAI,CAAC,IAAI,CAAC,IAAI,CAAC,GAAG,CAAC,IAAI,CAAC,IAAI,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC,GAAG,CAAC,CAAC;IAC7C,OAAO,WAAW,CAAC,IAAI,EAAE,WAAW,CAAC,CAAC;AACxC,CAAC;AAED,SAAS,aAAa,CAAC,IAAU,EAAE,MAAc;IAC/C,MAAM,GAAG,GAAG,IAAI,CAAC,OAAO,CAAC,cAAc,CAAC;IACxC,IAAI,CAAC,GAAG;QAAE,OAAO,SAAS,CAAC;IAC3B,OAAO,UAAU,CAAC,QAAQ,EAAE,MAAM,CAAC,IAAI,CAAC,GAAG,EAAE,KAAK,CAAC,CAAC,CAAC,MAAM,CAAC,MAAM,CAAC,CAAC,MAAM,CAAC,KAAK,CAAC,CAAC,KAAK,CAAC,CAAC,EAAE,sBAAsB,CAAC,CAAC;AACrH,CAAC;AAED,sGAAsG;AACtG,SAAS,iBAAiB,CAAC,KAAa;IACtC,MAAM,OAAO,GAAG,KAAK,CAAC,IAAI,EAAE,CAAC;IAC7B,OAAO,CACL,OAAO,CAAC,MAAM,IAAI,qBAAqB;QACvC,CAAC,IAAI,CAAC,IAAI,CAAC,OAAO,CAAC;QACnB,CAAC,aAAa,CAAC,IAAI,CAAC,OAAO,CAAC;QAC5B,8EAA8E;QAC9E,CAAC,OAAO,CAAC,QAAQ,CAAC,kBAAkB,CAAC,CACtC,CAAC;AACJ,CAAC;AAED,SAAS,aAAa,CAAC,IAAY,EAAE,IAAU;IAC7C,IAAI,GAAG,GAAG,IAAI,CAAC;IACf,KAAK,MAAM,MAAM,IAAI,cAAc,EAAE,CAAC;QACpC,GAAG,GAAG,GAAG,CAAC,OAAO,CAAC,MAAM,CAAC,OAAO,EAAE,CAAC,KAAa,EAAE,MAAe,EAAE,EAAE;YACnE,MAAM,WAAW,GAAG,MAAM,CAAC,WAAW,CAAC,CAAC,CAAC,aAAa,CAAC,IAAI,EAAE,KAAK,CAAC,CAAC,CAAC,CAAC,SAAS,CAAC;YAChF,OAAO,GAAG,OAAO,MAAM,KAAK,QAAQ,CAAC,CAAC,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,GAAG,QAAQ,CAAC,IAAI,EAAE,MAAM,CAAC,IAAI,EAAE,WAAW,CAAC,EAAE,CAAC;QAClG,CAAC,CAAC,CAAC;IACL,CAAC;IACD,OAAO,GAAG,CAAC;AACb,CAAC;AAED,SAAS,iBAAiB,CAAC,IAAY,EAAE,IAAU;IACjD,OAAO,IAAI,CAAC,OAAO,CAAC,qBAAqB,EAAE,CAAC,KAAa,EAAE,IAAY,EAAE,KAAa,EAAE,EAAE,CACxF,iBAAiB,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC,GAAG,IAAI,GAAG,QAAQ,CAAC,IAAI,EAAE,oBAAoB,CAAC,EAAE,CAAC,CAAC,CAAC,KAAK,CACpF,CAAC;AACJ,CAAC;AAED,SAAS,kBAAkB,CAAC,IAAY,EAAE,IAAU;IAClD,mGAAmG;IACnG,MAAM,UAAU,GAAG,IAAI,CAAC,OAAO,CAAC,KAAK,EAAE,CAAC,KAAa,EAAE,MAAc,EAAE,EAAE,CACvE,KAAK,CAAC,UAAU,CAAC,MAAM,CAAC,IAAI,qBAAqB,CAAC,IAAI,CAAC,MAAM,CAAC;QAC5D,CAAC,CAAC,KAAK;QACP,CAAC,CAAC,GAAG,QAAQ,CAAC,IAAI,EAAE,OAAO,CAAC,IAAI,MAAM,EAAE,CAC3C,CAAC;IACF,OAAO,UAAU,CAAC,OAAO,CAAC,cAAc,EAAE,CAAC,MAAc,EAAE,IAAY,EAAE,EAAE,CAAC,GAAG,IAAI,GAAG,QAAQ,CAAC,IAAI,EAAE,UAAU,CAAC,EAAE,CAAC,CAAC;AACtH,CAAC;AAED,SAAS,WAAW,CAAC,IAAY,EAAE,IAAU;IAC3C,IAAI,CAAC,IAAI;QAAE,OAAO,IAAI,CAAC;IACvB,MAAM,GAAG,GAAG,iBAAiB,CAAC,aAAa,CAAC,IAAI,EAAE,IAAI,CAAC,EAAE,IAAI,CAAC,CAAC;IAC/D,OAAO,IAAI,CAAC,OAAO,CAAC,YAAY,CAAC,CAAC,CAAC,kBAAkB,CAAC,GAAG,EAAE,IAAI,CAAC,CAAC,CAAC,CAAC,GAAG,CAAC;AACzE,CAAC;AAED,SAAS,SAAS,CAAC,KAAa,EAAE,IAAU,EAAE,GAAuB;IACnE,IAAI,GAAG,KAAK,SAAS,IAAI,cAAc,CAAC,IAAI,CAAC,GAAG,CAAC,IAAI,iBAAiB,CAAC,KAAK,CAAC,EAAE,CAAC;QAC9E,kGAAkG;QAClG,MAAM,SAAS,GAAG,aAAa,CAAC,KAAK,EAAE,IAAI,CAAC,CAAC;QAC7C,OAAO,SAAS,KAAK,KAAK,CAAC,CAAC,CAAC,SAAS,CAAC,CAAC,CAAC,QAAQ,CAAC,IAAI,EAAE,oBAAoB,CAAC,CAAC;IAChF,CAAC;IACD,OAAO,WAAW,CAAC,KAAK,EAAE,IAAI,CAAC,CAAC;AAClC,CAAC;AAED,SAAS,aAAa,CAAC,KAAc;IACnC,OAAO,OAAO,CAAC,KAAK,CAAC,IAAI,OAAO,KAAK,KAAK,QAAQ,IAAI,CAAC,KAAK,CAAC,OAAO,CAAC,KAAK,CAAC,CAAC;AAC9E,CAAC;AAED;;;;;;GAMG;AACH,SAAS,gBAAgB,CAAC,KAAa,EAAE,IAAY,EAAE,YAAoB,EAAE,IAAU;IACrF,IAAI,CAAC,KAAK,CAAC,IAAI,EAAE,IAAI,kBAAkB,CAAC,IAAI,CAAC,KAAK,CAAC,IAAI,aAAa,CAAC,IAAI,CAAC,KAAK,CAAC,IAAI,EAAE,CAAC;QAAE,OAAO,KAAK,CAAC;IACtG,IAAI,KAAK,CAAC,QAAQ,CAAC,kBAAkB,CAAC;QAAE,OAAO,KAAK,CAAC;IACrD,MAAM,SAAS,GAAG,aAAa,CAAC,KAAK,EAAE,IAAI,CAAC,CAAC;IAC7C,IAAI,SAAS,KAAK,KAAK;QAAE,OAAO,SAAS,CAAC;IAC1C,OAAO,QAAQ,CAAC,IAAI,EAAE,cAAc,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC,oBAAoB,CAAC,CAAC,CAAC,YAAY,CAAC,CAAC;AACzF,CAAC;AAED,SAAS,cAAc,CAAC,KAA8B,EAAE,YAAoB,EAAE,IAAU,EAAE,KAAa;IACrG,MAAM,GAAG,GAA4B,EAAE,CAAC;IACxC,KAAK,MAAM,CAAC,IAAI,EAAE,CAAC,CAAC,IAAI,MAAM,CAAC,OAAO,CAAC,KAAK,CAAC,EAAE,CAAC;QAC9C,GAAG,CAAC,IAAI,CAAC,GAAG,OAAO,CAAC,KAAK,QAAQ,CAAC,CAAC,CAAC,gBAAgB,CAAC,CAAC,EAAE,IAAI,EAAE,YAAY,EAAE,IAAI,CAAC,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC,EAAE,IAAI,EAAE,IAAI,EAAE,KAAK,GAAG,CAAC,CAAC,CAAC;IACrH,CAAC;IACD,OAAO,GAAG,CAAC;AACb,CAAC;AAED,yFAAyF;AACzF,SAAS,QAAQ,CAAC,IAAe,EAAE,IAAU,EAAE,KAAa;IAC1D,OAAO,IAAI,CAAC,GAAG,CAAC,CAAC,GAAG,EAAE,CAAC,EAAE,EAAE;QACzB,MAAM,IAAI,GAAG,OAAO,IAAI,CAAC,CAAC,GAAG,CAAC,CAAC,KAAK,QAAQ,CAAC,CAAC,CAAC,eAAe,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC,GAAG,CAAC,CAAW,CAAC,CAAC,CAAC,CAAC,IAAI,CAAC;QAClG,IAAI,IAAI,IAAI,cAAc,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC,IAAI,OAAO,GAAG,KAAK,QAAQ,IAAI,iBAAiB,CAAC,GAAG,CAAC,EAAE,CAAC;YAC9F,MAAM,SAAS,GAAG,aAAa,CAAC,GAAG,EAAE,IAAI,CAAC,CAAC;YAC3C,OAAO,SAAS,KAAK,GAAG,CAAC,CAAC,CAAC,SAAS,CAAC,CAAC,CAAC,QAAQ,CAAC,IAAI,EAAE,oBAAoB,CAAC,CAAC;QAC9E,CAAC;QACD,OAAO,IAAI,CAAC,GAAG,EAAE,IAAI,EAAE,MAAM,EAAE,KAAK,GAAG,CAAC,CAAC,CAAC;IAC5C,CAAC,CAAC,CAAC;AACL,CAAC;AAED,uFAAuF;AACvF,SAAS,SAAS,CAAC,GAAW,EAAE,KAAc,EAAE,QAAiB,EAAE,IAAU,EAAE,KAAa;IAC1F,MAAM,SAAS,GAAG,QAAQ,CAAC,CAAC,CAAC,gBAAgB,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,SAAS,CAAC;IAC/D,IAAI,SAAS,IAAI,aAAa,CAAC,KAAK,CAAC;QAAE,OAAO,cAAc,CAAC,KAAK,EAAE,SAAS,EAAE,IAAI,EAAE,KAAK,CAAC,CAAC;IAC5F,IAAI,GAAG,KAAK,MAAM,IAAI,KAAK,CAAC,OAAO,CAAC,KAAK,CAAC;QAAE,OAAO,QAAQ,CAAC,KAAK,EAAE,IAAI,EAAE,KAAK,CAAC,CAAC;IAChF,OAAO,IAAI,CAAC,KAAK,EAAE,IAAI,EAAE,GAAG,EAAE,KAAK,GAAG,CAAC,CAAC,CAAC;AAC3C,CAAC;AAED,SAAS,UAAU,CAAC,GAA4B,EAAE,IAAU,EAAE,KAAa;IACzE,MAAM,QAAQ,GAAG,MAAM,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC,IAAI,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,eAAe,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,CAAC;IACtE,MAAM,GAAG,GAA4B,EAAE,CAAC;IACxC,KAAK,MAAM,CAAC,CAAC,EAAE,CAAC,CAAC,IAAI,MAAM,CAAC,OAAO,CAAC,GAAG,CAAC;QAAE,GAAG,CAAC,CAAC,CAAC,GAAG,SAAS,CAAC,CAAC,EAAE,CAAC,EAAE,QAAQ,EAAE,IAAI,EAAE,KAAK,CAAC,CAAC;IAC1F,OAAO,GAAG,CAAC;AACb,CAAC;AAED,SAAS,IAAI,CAAC,KAAc,EAAE,IAAU,EAAE,GAAuB,EAAE,KAAa;IAC9E,IAAI,KAAK,GAAG,SAAS;QAAE,OAAO,QAAQ,CAAC,IAAI,EAAE,UAAU,CAAC,CAAC;IACzD,IAAI,KAAK,CAAC,OAAO,CAAC,KAAK,CAAC;QAAE,OAAO,KAAK,CAAC,GAAG,CAAC,CAAC,IAAI,EAAE,EAAE,CAAC,IAAI,CAAC,IAAI,EAAE,IAAI,EAAE,GAAG,EAAE,KAAK,GAAG,CAAC,CAAC,CAAC,CAAC;IACvF,IAAI,aAAa,CAAC,KAAK,CAAC;QAAE,OAAO,UAAU,CAAC,KAAK,EAAE,IAAI,EAAE,KAAK,CAAC,CAAC;IAChE,OAAO,OAAO,KAAK,KAAK,QAAQ,CAAC,CAAC,CAAC,SAAS,CAAC,KAAK,EAAE,IAAI,EAAE,GAAG,CAAC,CAAC,CAAC,CAAC,KAAK,CAAC;AACzE,CAAC;AAED;;;GAGG;AACH,MAAM,UAAU,KAAK,CAAI,KAAQ,EAAE,UAAwB,EAAE;IAC3D,MAAM,IAAI,GAAS,EAAE,OAAO,EAAE,IAAI,EAAE,EAAE,EAAE,CAAC;IACzC,OAAO,EAAE,KAAK,EAAE,IAAI,CAAC,KAAK,EAAE,IAAI,EAAE,SAAS,EAAE,CAAC,CAAM,EAAE,IAAI,EAAE,IAAI,CAAC,IAAI,EAAE,CAAC;AAC1E,CAAC;AAED,4BAA4B;AAC5B,MAAM,UAAU,UAAU,CAAI,KAAQ,EAAE,UAAwB,EAAE;IAChE,OAAO,KAAK,CAAC,KAAK,EAAE,OAAO,CAAC,CAAC,KAAK,CAAC;AACrC,CAAC;AAED,kGAAkG;AAClG,MAAM,UAAU,SAAS,CAAC,IAAY,EAAE,UAAwB,EAAE;IAChE,OAAO,WAAW,CAAC,IAAI,EAAE,EAAE,OAAO,EAAE,IAAI,EAAE,EAAE,EAAE,CAAC,CAAC;AAClD,CAAC;AAED;;;GAGG;AACH,MAAM,UAAU,WAAW,CACzB,MAAS,EACT,MAAyB,EACzB,UAAwB,EAAE;IAE1B,MAAM,GAAG,GAA4B,EAAE,GAAG,MAAM,EAAE,CAAC;IACnD,KAAK,MAAM,KAAK,IAAI,MAAM,EAAE,CAAC;QAC3B,IAAI,OAAO,GAAG,CAAC,KAAK,CAAC,KAAK,QAAQ;YAAE,GAAG,CAAC,KAAK,CAAC,GAAG,SAAS,CAAC,GAAG,CAAC,KAAK,CAAW,EAAE,OAAO,CAAC,CAAC;IAC5F,CAAC;IACD,OAAO,GAAQ,CAAC;AAClB,CAAC;AAED,mGAAmG;AACnG,MAAM,UAAU,mBAAmB,CAAC,KAAc;IAChD,IAAI,OAAO,KAAK,KAAK,QAAQ;QAAE,OAAO,KAAK,CAAC,QAAQ,CAAC,kBAAkB,CAAC,CAAC;IACzE,IAAI,CAAC;QACH,OAAO,IAAI,CAAC,SAAS,CAAC,KAAK,IAAI,IAAI,CAAC,CAAC,QAAQ,CAAC,kBAAkB,CAAC,CAAC;IACpE,CAAC;IAAC,MAAM,CAAC;QACP,OAAO,KAAK,CAAC;IACf,CAAC;AACH,CAAC"}

package/dist/rules.d.ts

@@ -0,0 +1,32 @@
+/**
+ * The scrubber's rules, loaded from `rules.json` at the package root.
+ *
+ * The rules are data, checked in, so the backend's safety net reads the very same file (and a
+ * Python test compiles every pattern in it). Each pattern is written for both the JavaScript and
+ * Python regex engines: no lookbehind, no named groups, no Unicode property escapes, and every
+ * repeat that can start a match is bounded, so a long run of letters scans in linear time.
+ *
+ * Formats run in file order, most specific first (Anthropic before the generic `sk-` OpenAI
+ * shape, vendor tokens before `Bearer`), so a value is typed by the most precise rule that
+ * matches. Group 1 of a format, when present, is the published vendor prefix that stays.
+ */
+export interface SecretFormat {
+    type: string;
+    fingerprint: boolean;
+    /** Global, for `replace`. */
+    pattern: RegExp;
+}
+export declare const SECRET_FORMATS: readonly SecretFormat[];
+export declare const CREDENTIAL_KEY: RegExp;
+export declare const CREDENTIAL_ASSIGNMENT: RegExp;
+export declare const ENV_REFERENCE: RegExp;
+export declare const CONTAINS_REFERENCE: RegExp;
+export declare const CREDENTIAL_FLAG: RegExp;
+/** Block key (`env`, `headers`) to the placeholder type for an ordinary value in it. */
+export declare const MCP_VALUE_BLOCKS: Readonly<Record<string, string>>;
+export declare const MCP_SERVER_KEYS: ReadonlySet<string>;
+export declare const EMAIL: RegExp;
+export declare const RESERVED_EMAIL_DOMAIN: RegExp;
+export declare const HOME_DIRECTORY: RegExp;
+export declare const MIN_CREDENTIAL_LENGTH: number;
+//# sourceMappingURL=rules.d.ts.map

package/dist/rules.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"rules.d.ts","sourceRoot":"","sources":["../src/rules.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;GAWG;AA6BH,MAAM,WAAW,YAAY;IAC3B,IAAI,EAAE,MAAM,CAAC;IACb,WAAW,EAAE,OAAO,CAAC;IACrB,6BAA6B;IAC7B,OAAO,EAAE,MAAM,CAAC;CACjB;AAWD,eAAO,MAAM,cAAc,EAAE,SAAS,YAAY,EAI/C,CAAC;AACJ,eAAO,MAAM,cAAc,QAAuC,CAAC;AACnE,eAAO,MAAM,qBAAqB,QAA6C,CAAC;AAChF,eAAO,MAAM,aAAa,QAAsC,CAAC;AACjE,eAAO,MAAM,kBAAkB,QAA2C,CAAC;AAC3E,eAAO,MAAM,eAAe,QAAwC,CAAC;AACrE,wFAAwF;AACxF,eAAO,MAAM,gBAAgB,EAAE,QAAQ,CAAC,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC,CAAiC,CAAC;AAChG,eAAO,MAAM,eAAe,EAAE,WAAW,CAAC,MAAM,CAA+C,CAAC;AAChG,eAAO,MAAM,KAAK,QAA6B,CAAC;AAChD,eAAO,MAAM,qBAAqB,QAA8C,CAAC;AACjF,eAAO,MAAM,cAAc,QAAsC,CAAC;AAClE,eAAO,MAAM,qBAAqB,QAA8B,CAAC"}

package/dist/rules.js

@@ -0,0 +1,38 @@
+/**
+ * The scrubber's rules, loaded from `rules.json` at the package root.
+ *
+ * The rules are data, checked in, so the backend's safety net reads the very same file (and a
+ * Python test compiles every pattern in it). Each pattern is written for both the JavaScript and
+ * Python regex engines: no lookbehind, no named groups, no Unicode property escapes, and every
+ * repeat that can start a match is bounded, so a long run of letters scans in linear time.
+ *
+ * Formats run in file order, most specific first (Anthropic before the generic `sk-` OpenAI
+ * shape, vendor tokens before `Bearer`), so a value is typed by the most precise rule that
+ * matches. Group 1 of a format, when present, is the published vendor prefix that stays.
+ */
+import { createRequire } from "node:module";
+// dist/rules.js → ../rules.json, the file at the package root (shipped in `files`). Loaded with
+// Node's JSON loader rather than `fs`, which a consumer's tests may mock.
+const RULES = createRequire(import.meta.url)("../rules.json");
+/** `global` for patterns used with `replace`; never for `test`, whose `lastIndex` would stick. */
+function compile(spec, global) {
+    return new RegExp(spec.pattern, `${global ? "g" : ""}${spec.ignore_case ? "i" : ""}`);
+}
+export const SECRET_FORMATS = RULES.formats.map((f) => ({
+    type: f.type,
+    fingerprint: f.fingerprint,
+    pattern: compile(f, true),
+}));
+export const CREDENTIAL_KEY = compile(RULES.credential_key, false);
+export const CREDENTIAL_ASSIGNMENT = compile(RULES.credential_assignment, true);
+export const ENV_REFERENCE = compile(RULES.env_reference, false);
+export const CONTAINS_REFERENCE = compile(RULES.contains_reference, false);
+export const CREDENTIAL_FLAG = compile(RULES.credential_flag, false);
+/** Block key (`env`, `headers`) to the placeholder type for an ordinary value in it. */
+export const MCP_VALUE_BLOCKS = RULES.mcp_value_blocks.blocks;
+export const MCP_SERVER_KEYS = new Set(RULES.mcp_value_blocks.server_keys);
+export const EMAIL = compile(RULES.email, true);
+export const RESERVED_EMAIL_DOMAIN = compile(RULES.reserved_email_domain, false);
+export const HOME_DIRECTORY = compile(RULES.home_directory, true);
+export const MIN_CREDENTIAL_LENGTH = RULES.min_credential_length;
+//# sourceMappingURL=rules.js.map

package/dist/rules.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"rules.js","sourceRoot":"","sources":["../src/rules.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;GAWG;AAEH,OAAO,EAAE,aAAa,EAAE,MAAM,aAAa,CAAC;AAkC5C,gGAAgG;AAChG,0EAA0E;AAC1E,MAAM,KAAK,GAAG,aAAa,CAAC,MAAM,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC,eAAe,CAAc,CAAC;AAE3E,kGAAkG;AAClG,SAAS,OAAO,CAAC,IAAiB,EAAE,MAAe;IACjD,OAAO,IAAI,MAAM,CAAC,IAAI,CAAC,OAAO,EAAE,GAAG,MAAM,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,GAAG,IAAI,CAAC,WAAW,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC;AACxF,CAAC;AAED,MAAM,CAAC,MAAM,cAAc,GAA4B,KAAK,CAAC,OAAO,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC;IAC/E,IAAI,EAAE,CAAC,CAAC,IAAI;IACZ,WAAW,EAAE,CAAC,CAAC,WAAW;IAC1B,OAAO,EAAE,OAAO,CAAC,CAAC,EAAE,IAAI,CAAC;CAC1B,CAAC,CAAC,CAAC;AACJ,MAAM,CAAC,MAAM,cAAc,GAAG,OAAO,CAAC,KAAK,CAAC,cAAc,EAAE,KAAK,CAAC,CAAC;AACnE,MAAM,CAAC,MAAM,qBAAqB,GAAG,OAAO,CAAC,KAAK,CAAC,qBAAqB,EAAE,IAAI,CAAC,CAAC;AAChF,MAAM,CAAC,MAAM,aAAa,GAAG,OAAO,CAAC,KAAK,CAAC,aAAa,EAAE,KAAK,CAAC,CAAC;AACjE,MAAM,CAAC,MAAM,kBAAkB,GAAG,OAAO,CAAC,KAAK,CAAC,kBAAkB,EAAE,KAAK,CAAC,CAAC;AAC3E,MAAM,CAAC,MAAM,eAAe,GAAG,OAAO,CAAC,KAAK,CAAC,eAAe,EAAE,KAAK,CAAC,CAAC;AACrE,wFAAwF;AACxF,MAAM,CAAC,MAAM,gBAAgB,GAAqC,KAAK,CAAC,gBAAgB,CAAC,MAAM,CAAC;AAChG,MAAM,CAAC,MAAM,eAAe,GAAwB,IAAI,GAAG,CAAC,KAAK,CAAC,gBAAgB,CAAC,WAAW,CAAC,CAAC;AAChG,MAAM,CAAC,MAAM,KAAK,GAAG,OAAO,CAAC,KAAK,CAAC,KAAK,EAAE,IAAI,CAAC,CAAC;AAChD,MAAM,CAAC,MAAM,qBAAqB,GAAG,OAAO,CAAC,KAAK,CAAC,qBAAqB,EAAE,KAAK,CAAC,CAAC;AACjF,MAAM,CAAC,MAAM,cAAc,GAAG,OAAO,CAAC,KAAK,CAAC,cAAc,EAAE,IAAI,CAAC,CAAC;AAClE,MAAM,CAAC,MAAM,qBAAqB,GAAG,KAAK,CAAC,qBAAqB,CAAC"}

package/package.json

@@ -1,6 +1,32 @@
 {
   "name": "@optimuslabs/pii-scrub-staging",
-  "version": "0.0.0-stage",
-  "stub": true,
-  "description": "Temporary package placeholder for staged publishing"
-}
+  "version": "0.1.0",
+  "description": "Redacts secrets and personal data from what an Optimus endpoint sends, before it leaves the machine.",
+  "type": "module",
+  "main": "./dist/index.js",
+  "types": "./dist/index.d.ts",
+  "files": [
+    "dist/**/*",
+    "!dist/test/**",
+    "rules.json",
+    "README.md"
+  ],
+  "scripts": {
+    "build": "rm -rf dist && tsc -p tsconfig.json",
+    "pretest": "npm run build",
+    "test": "node --test dist/test/*.test.js",
+    "prepack": "npm run build"
+  },
+  "engines": {
+    "node": ">=18"
+  },
+  "license": "UNLICENSED",
+  "dependencies": {},
+  "devDependencies": {
+    "@types/node": "^20.19.39",
+    "typescript": "^5.4.5"
+  },
+  "publishConfig": {
+    "access": "public"
+  }
+}

package/rules.json

@@ -0,0 +1,154 @@
+{
+  "description": "Rules for @optimuslabs/pii-scrub, read by the endpoint scrubber and the backend. Every pattern is valid in both the JavaScript and Python regex engines; see docs/pii-scrubbing/DESIGN.md.",
+  "placeholder": {
+    "description": "A placeholder the scrubber wrote: left unchanged and never counted as a hit.",
+    "pattern": "\\[OL_REDACTED:[A-Z_]+(?::[0-9a-f]{24})?\\]"
+  },
+  "formats": [
+    {
+      "type": "PRIVATE_KEY",
+      "pattern": "-----BEGIN (?:[A-Z0-9]+ )*PRIVATE KEY-----[\\s\\S]*?-----END (?:[A-Z0-9]+ )*PRIVATE KEY-----",
+      "fingerprint": true
+    },
+    {
+      "type": "URL_CREDENTIALS",
+      "pattern": "\\b([a-z][a-z0-9+.-]{0,31}:\\/\\/[^\\s:@/\"'`]{1,256}:)[^\\s@/\"'`]{1,256}(?=@)",
+      "ignore_case": true,
+      "fingerprint": false
+    },
+    {
+      "type": "ANTHROPIC_KEY",
+      "pattern": "\\b(sk-ant-[a-z]+\\d{2}-)[A-Za-z0-9_-]{20,}",
+      "fingerprint": true
+    },
+    {
+      "type": "OPENAI_KEY",
+      "pattern": "\\b(sk-(?:proj|svcacct|admin)-)[A-Za-z0-9_-]{20,}",
+      "fingerprint": true
+    },
+    {
+      "type": "OPENAI_KEY",
+      "pattern": "\\b(sk-)[A-Za-z0-9]{20,}",
+      "fingerprint": true
+    },
+    {
+      "type": "GITHUB_TOKEN",
+      "pattern": "\\b(gh[pousr]_)[A-Za-z0-9]{36,}",
+      "fingerprint": true
+    },
+    {
+      "type": "GITHUB_TOKEN",
+      "pattern": "\\b(github_pat_)[A-Za-z0-9_]{22,}",
+      "fingerprint": true
+    },
+    {
+      "type": "AWS_ACCESS_KEY",
+      "pattern": "\\b(AKIA|ASIA)[0-9A-Z]{16}\\b",
+      "fingerprint": true
+    },
+    {
+      "type": "STRIPE_KEY",
+      "pattern": "\\b((?:sk|rk)_(?:live|test)_)[A-Za-z0-9]{16,}",
+      "fingerprint": true
+    },
+    {
+      "type": "SLACK_TOKEN",
+      "pattern": "\\b(xox[abposr]-|xapp-)[A-Za-z0-9-]{10,}",
+      "fingerprint": true
+    },
+    {
+      "type": "GOOGLE_API_KEY",
+      "pattern": "\\b(AIza)[0-9A-Za-z_-]{35}",
+      "fingerprint": true
+    },
+    {
+      "type": "NPM_TOKEN",
+      "pattern": "\\b(npm_)[A-Za-z0-9]{36}\\b",
+      "fingerprint": true
+    },
+    {
+      "type": "PYPI_TOKEN",
+      "pattern": "\\b(pypi-)AgEIcHlwaS5vcmc[A-Za-z0-9_-]{50,}",
+      "fingerprint": true
+    },
+    {
+      "type": "SENDGRID_KEY",
+      "pattern": "\\b(SG\\.)[A-Za-z0-9_-]{22}\\.[A-Za-z0-9_-]{43}",
+      "fingerprint": true
+    },
+    {
+      "type": "HUGGINGFACE_TOKEN",
+      "pattern": "\\b(hf_)[A-Za-z0-9]{34,}",
+      "fingerprint": true
+    },
+    {
+      "type": "PERPLEXITY_KEY",
+      "pattern": "\\b(pplx-)[A-Za-z0-9]{32,}",
+      "fingerprint": true
+    },
+    {
+      "type": "JWT",
+      "pattern": "\\beyJ[A-Za-z0-9_-]{8,}\\.eyJ[A-Za-z0-9_-]{8,}\\.[A-Za-z0-9_-]{8,}",
+      "fingerprint": true
+    },
+    {
+      "type": "BEARER_TOKEN",
+      "pattern": "\\b(Bearer\\s+)[A-Za-z0-9._~+/=-]{16,}",
+      "ignore_case": true,
+      "fingerprint": false
+    },
+    {
+      "type": "BASIC_AUTH",
+      "pattern": "\\b(Basic\\s+)[A-Za-z0-9+/=]{12,}",
+      "fingerprint": false
+    }
+  ],
+  "credential_key": {
+    "pattern": "(?:[_.-]key|api[_-]?key|apikey|secret|token|password|passwd|credential|private[_-]?key|access[_-]?key|authorization|token[_-]?cache(?:v\\d+)?)$",
+    "ignore_case": true
+  },
+  "credential_assignment": {
+    "pattern": "((?:api[_-]?key|apikey|secret|token|password|passwd|credential|private[_-]?key|access[_-]?key)[\"']?\\s*[:=]\\s*[\"']?)([^\\s\"'`,;&)}\\]]{8,})",
+    "ignore_case": true
+  },
+  "env_reference": {
+    "pattern": "^(?:\\$|\\{\\{|process\\.env\\.)|^[A-Z][A-Z0-9]*(?:_[A-Z0-9]+)+$"
+  },
+  "contains_reference": {
+    "description": "A reference to a secret stored elsewhere anywhere in a value (`Bearer ${TOKEN}`): the value stays readable.",
+    "pattern": "\\$\\{[^}]{1,128}\\}|\\$env:|process\\.env\\."
+  },
+  "credential_flag": {
+    "description": "A command-line flag whose next argument is a credential (`--api-key`, `--token`); group 1 is the name, tested against credential_key.",
+    "pattern": "^-{1,2}([A-Za-z0-9_.-]{1,64})$"
+  },
+  "mcp_value_blocks": {
+    "description": "Blocks inside an MCP server definition whose every value is redacted: block key to the placeholder type for a value that is neither a known format nor under a credential-named key. A server definition is an object with one of server_keys.",
+    "blocks": {
+      "env": "ENV_VALUE",
+      "environment": "ENV_VALUE",
+      "headers": "HEADER_VALUE",
+      "http_headers": "HEADER_VALUE"
+    },
+    "server_keys": [
+      "command",
+      "args",
+      "url",
+      "serverUrl",
+      "httpUrl",
+      "transport",
+      "type"
+    ]
+  },
+  "email": {
+    "pattern": "\\b[A-Za-z0-9._%+-]{1,64}@([A-Za-z0-9-]{1,63}(?:\\.[A-Za-z0-9-]{1,63})*\\.[A-Za-z]{2,24})\\b"
+  },
+  "reserved_email_domain": {
+    "pattern": "(?:^|\\.)(?:example\\.(?:com|org|net)|example|test|invalid|localhost)$",
+    "ignore_case": true
+  },
+  "home_directory": {
+    "pattern": "(\\/(?:Users|home)\\/)([A-Za-z0-9._-]{1,64})"
+  },
+  "min_credential_length": 8
+}