@owasp-aghast/aghast 0.8.3 → 0.9.0-rc.1

package/dist/cost-tracker.d.ts

@@ -0,0 +1,45 @@
+/**
+ * Shared cost-tracking utilities for the scan runner and judge stage.
+ */
+import { type PricingConfig, type CostBreakdown } from './cost-calculator.js';
+import { type BudgetLimits } from './budget.js';
+import type { TokenUsage } from './types.js';
+import type { ScanRecord } from './scan-history.js';
+/**
+ * Tracks accumulated tokens/cost across a scan so the budget can be evaluated
+ * before each AI call. Mutated in place by AI invocations.
+ */
+export interface ScanCostTracker {
+    pricing?: PricingConfig;
+    budgetLimits?: BudgetLimits;
+    scanHistory?: ScanRecord[];
+    totalTokens: number;
+    totalCostUsd: number;
+    /** Cost source from the last recorded AI call. Used for banner labelling. */
+    lastCostSource?: CostBreakdown['source'];
+    lastCostReportedBy?: CostBreakdown['reportedBy'];
+    lastCostCoveredBySubscription?: boolean;
+    /** Set true after the first warn so we don't log it repeatedly. */
+    warned: boolean;
+    /** Most recent budget action returned to the runner. */
+    lastAction: 'continue' | 'warn' | 'abort';
+    /** Reason from the most recent non-continue check. */
+    lastReason?: string;
+}
+export declare function createCostTracker(options: {
+    pricing?: PricingConfig;
+    budgetLimits?: BudgetLimits;
+    scanHistory?: ScanRecord[];
+}): ScanCostTracker;
+/**
+ * Record an AI call's token usage against the tracker. Called after each
+ * successful executeCheck().
+ */
+export declare function recordUsage(tracker: ScanCostTracker, usage: TokenUsage | undefined, model: string): void;
+/**
+ * Check the budget before an AI call. Logs a warning the first time the warn
+ * threshold is crossed; throws BudgetExceededError when the abort threshold is
+ * crossed.
+ */
+export declare function preflightBudget(tracker: ScanCostTracker): void;
+//# sourceMappingURL=cost-tracker.d.ts.map

package/dist/cost-tracker.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"cost-tracker.d.ts","sourceRoot":"","sources":["../src/cost-tracker.ts"],"names":[],"mappings":"AAAA;;GAEG;AAEH,OAAO,EAAiB,KAAK,aAAa,EAAE,KAAK,aAAa,EAAE,MAAM,sBAAsB,CAAC;AAC7F,OAAO,EAAoC,KAAK,YAAY,EAAE,MAAM,aAAa,CAAC;AAElF,OAAO,KAAK,EAAE,UAAU,EAAE,MAAM,YAAY,CAAC;AAC7C,OAAO,KAAK,EAAE,UAAU,EAAE,MAAM,mBAAmB,CAAC;AAIpD;;;GAGG;AACH,MAAM,WAAW,eAAe;IAC9B,OAAO,CAAC,EAAE,aAAa,CAAC;IACxB,YAAY,CAAC,EAAE,YAAY,CAAC;IAC5B,WAAW,CAAC,EAAE,UAAU,EAAE,CAAC;IAC3B,WAAW,EAAE,MAAM,CAAC;IACpB,YAAY,EAAE,MAAM,CAAC;IACrB,6EAA6E;IAC7E,cAAc,CAAC,EAAE,aAAa,CAAC,QAAQ,CAAC,CAAC;IACzC,kBAAkB,CAAC,EAAE,aAAa,CAAC,YAAY,CAAC,CAAC;IACjD,6BAA6B,CAAC,EAAE,OAAO,CAAC;IACxC,mEAAmE;IACnE,MAAM,EAAE,OAAO,CAAC;IAChB,wDAAwD;IACxD,UAAU,EAAE,UAAU,GAAG,MAAM,GAAG,OAAO,CAAC;IAC1C,sDAAsD;IACtD,UAAU,CAAC,EAAE,MAAM,CAAC;CACrB;AAED,wBAAgB,iBAAiB,CAAC,OAAO,EAAE;IACzC,OAAO,CAAC,EAAE,aAAa,CAAC;IACxB,YAAY,CAAC,EAAE,YAAY,CAAC;IAC5B,WAAW,CAAC,EAAE,UAAU,EAAE,CAAC;CAC5B,GAAG,eAAe,CAUlB;AAED;;;GAGG;AACH,wBAAgB,WAAW,CACzB,OAAO,EAAE,eAAe,EACxB,KAAK,EAAE,UAAU,GAAG,SAAS,EAC7B,KAAK,EAAE,MAAM,GACZ,IAAI,CAQN;AAED;;;;GAIG;AACH,wBAAgB,eAAe,CAAC,OAAO,EAAE,eAAe,GAAG,IAAI,CAmB9D"}

package/dist/cost-tracker.js

@@ -0,0 +1,56 @@
+/**
+ * Shared cost-tracking utilities for the scan runner and judge stage.
+ */
+import { calculateCost } from './cost-calculator.js';
+import { checkBudget, BudgetExceededError } from './budget.js';
+import { logProgress } from './logging.js';
+const TAG = 'scan';
+export function createCostTracker(options) {
+    return {
+        pricing: options.pricing,
+        budgetLimits: options.budgetLimits,
+        scanHistory: options.scanHistory,
+        totalTokens: 0,
+        totalCostUsd: 0,
+        warned: false,
+        lastAction: 'continue',
+    };
+}
+/**
+ * Record an AI call's token usage against the tracker. Called after each
+ * successful executeCheck().
+ */
+export function recordUsage(tracker, usage, model) {
+    if (!usage || !tracker.pricing)
+        return;
+    const cost = calculateCost(usage, model, tracker.pricing);
+    tracker.totalTokens += usage.totalTokens;
+    tracker.totalCostUsd += cost.totalCost;
+    tracker.lastCostSource = cost.source;
+    tracker.lastCostReportedBy = cost.reportedBy;
+    tracker.lastCostCoveredBySubscription = cost.coveredBySubscription;
+}
+/**
+ * Check the budget before an AI call. Logs a warning the first time the warn
+ * threshold is crossed; throws BudgetExceededError when the abort threshold is
+ * crossed.
+ */
+export function preflightBudget(tracker) {
+    if (!tracker.budgetLimits)
+        return;
+    const status = checkBudget({
+        currentScanCostUsd: tracker.totalCostUsd,
+        currentScanTokens: tracker.totalTokens,
+        history: tracker.scanHistory,
+    }, tracker.budgetLimits);
+    tracker.lastAction = status.action;
+    tracker.lastReason = status.reason;
+    if (status.action === 'abort') {
+        throw new BudgetExceededError(status.reason ?? 'Budget limit exceeded');
+    }
+    if (status.action === 'warn' && !tracker.warned) {
+        tracker.warned = true;
+        logProgress(TAG, `Budget warning: ${status.reason ?? 'approaching limit'}`);
+    }
+}
+//# sourceMappingURL=cost-tracker.js.map

package/dist/cost-tracker.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"cost-tracker.js","sourceRoot":"","sources":["../src/cost-tracker.ts"],"names":[],"mappings":"AAAA;;GAEG;AAEH,OAAO,EAAE,aAAa,EAA0C,MAAM,sBAAsB,CAAC;AAC7F,OAAO,EAAE,WAAW,EAAE,mBAAmB,EAAqB,MAAM,aAAa,CAAC;AAClF,OAAO,EAAE,WAAW,EAAE,MAAM,cAAc,CAAC;AAI3C,MAAM,GAAG,GAAG,MAAM,CAAC;AAwBnB,MAAM,UAAU,iBAAiB,CAAC,OAIjC;IACC,OAAO;QACL,OAAO,EAAE,OAAO,CAAC,OAAO;QACxB,YAAY,EAAE,OAAO,CAAC,YAAY;QAClC,WAAW,EAAE,OAAO,CAAC,WAAW;QAChC,WAAW,EAAE,CAAC;QACd,YAAY,EAAE,CAAC;QACf,MAAM,EAAE,KAAK;QACb,UAAU,EAAE,UAAU;KACvB,CAAC;AACJ,CAAC;AAED;;;GAGG;AACH,MAAM,UAAU,WAAW,CACzB,OAAwB,EACxB,KAA6B,EAC7B,KAAa;IAEb,IAAI,CAAC,KAAK,IAAI,CAAC,OAAO,CAAC,OAAO;QAAE,OAAO;IACvC,MAAM,IAAI,GAAG,aAAa,CAAC,KAAK,EAAE,KAAK,EAAE,OAAO,CAAC,OAAO,CAAC,CAAC;IAC1D,OAAO,CAAC,WAAW,IAAI,KAAK,CAAC,WAAW,CAAC;IACzC,OAAO,CAAC,YAAY,IAAI,IAAI,CAAC,SAAS,CAAC;IACvC,OAAO,CAAC,cAAc,GAAG,IAAI,CAAC,MAAM,CAAC;IACrC,OAAO,CAAC,kBAAkB,GAAG,IAAI,CAAC,UAAU,CAAC;IAC7C,OAAO,CAAC,6BAA6B,GAAG,IAAI,CAAC,qBAAqB,CAAC;AACrE,CAAC;AAED;;;;GAIG;AACH,MAAM,UAAU,eAAe,CAAC,OAAwB;IACtD,IAAI,CAAC,OAAO,CAAC,YAAY;QAAE,OAAO;IAClC,MAAM,MAAM,GAAG,WAAW,CACxB;QACE,kBAAkB,EAAE,OAAO,CAAC,YAAY;QACxC,iBAAiB,EAAE,OAAO,CAAC,WAAW;QACtC,OAAO,EAAE,OAAO,CAAC,WAAW;KAC7B,EACD,OAAO,CAAC,YAAY,CACrB,CAAC;IACF,OAAO,CAAC,UAAU,GAAG,MAAM,CAAC,MAAM,CAAC;IACnC,OAAO,CAAC,UAAU,GAAG,MAAM,CAAC,MAAM,CAAC;IACnC,IAAI,MAAM,CAAC,MAAM,KAAK,OAAO,EAAE,CAAC;QAC9B,MAAM,IAAI,mBAAmB,CAAC,MAAM,CAAC,MAAM,IAAI,uBAAuB,CAAC,CAAC;IAC1E,CAAC;IACD,IAAI,MAAM,CAAC,MAAM,KAAK,MAAM,IAAI,CAAC,OAAO,CAAC,MAAM,EAAE,CAAC;QAChD,OAAO,CAAC,MAAM,GAAG,IAAI,CAAC;QACtB,WAAW,CAAC,GAAG,EAAE,mBAAmB,MAAM,CAAC,MAAM,IAAI,mBAAmB,EAAE,CAAC,CAAC;IAC9E,CAAC;AACH,CAAC"}

package/dist/discoveries/glob-discovery.d.ts

@@ -0,0 +1,21 @@
+/**
+ * Glob-based target discovery.
+ *
+ * Matches files in the repository by glob pattern (e.g. "src/routes/**\/*.ts")
+ * and returns each matched file as a whole-file target. Useful for checks that
+ * apply to entire files (e.g. "review every GraphQL resolver"), without
+ * requiring an external tool such as Semgrep.
+ *
+ * Spec reference: Appendix E.2.1.
+ */
+import type { TargetDiscovery, DiscoveredTarget } from '../discovery.js';
+/**
+ * Discover whole-file targets in `repoPath` matching `pattern`.
+ *
+ * Exposed as a named export (rather than only the registry-facing object) so
+ * unit tests can exercise the matching logic directly without going through
+ * the SecurityCheck plumbing.
+ */
+export declare function discoverGlobTargets(pattern: string, repoPath: string, maxTargets?: number): Promise<DiscoveredTarget[]>;
+export declare const globDiscovery: TargetDiscovery;
+//# sourceMappingURL=glob-discovery.d.ts.map

package/dist/discoveries/glob-discovery.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"glob-discovery.d.ts","sourceRoot":"","sources":["../../src/discoveries/glob-discovery.ts"],"names":[],"mappings":"AAAA;;;;;;;;;GASG;AASH,OAAO,KAAK,EAAE,eAAe,EAAE,gBAAgB,EAAoB,MAAM,iBAAiB,CAAC;AAmI3F;;;;;;GAMG;AACH,wBAAsB,mBAAmB,CACvC,OAAO,EAAE,MAAM,EACf,QAAQ,EAAE,MAAM,EAChB,UAAU,CAAC,EAAE,MAAM,GAClB,OAAO,CAAC,gBAAgB,EAAE,CAAC,CAoD7B;AAED,eAAO,MAAM,aAAa,EAAE,eAoC3B,CAAC"}

package/dist/discoveries/glob-discovery.js

@@ -0,0 +1,215 @@
+/**
+ * Glob-based target discovery.
+ *
+ * Matches files in the repository by glob pattern (e.g. "src/routes/**\/*.ts")
+ * and returns each matched file as a whole-file target. Useful for checks that
+ * apply to entire files (e.g. "review every GraphQL resolver"), without
+ * requiring an external tool such as Semgrep.
+ *
+ * Spec reference: Appendix E.2.1.
+ */
+import { readdir, stat } from 'node:fs/promises';
+import { createReadStream } from 'node:fs';
+import { join, relative, sep } from 'node:path';
+import picomatch from 'picomatch';
+import { logDebug } from '../logging.js';
+import { ERROR_CODES, formatError } from '../error-codes.js';
+import { DEFAULT_GENERIC_PROMPT } from '../defaults.js';
+const TAG = 'glob-discovery';
+/**
+ * Skip files larger than this (in bytes) to avoid OOM when a wide glob
+ * (e.g. "**\/*") accidentally matches a large generated/binary file.
+ * Files above this threshold are excluded from the target set with a debug log.
+ */
+const MAX_FILE_SIZE_BYTES = 10 * 1024 * 1024; // 10 MiB
+/** Directories that are always skipped when walking the repository tree. */
+const ALWAYS_SKIP_DIRS = new Set([
+    '.git',
+    'node_modules',
+    '.venv',
+    'venv',
+    '__pycache__',
+    '.tox',
+    '.mypy_cache',
+    '.pytest_cache',
+    'dist',
+    'build',
+    '.next',
+    '.nuxt',
+    '.cache',
+    '.idea',
+    '.vscode',
+]);
+function buildTargetPromptEnrichment(file, endLine) {
+    return `\n\nTARGET FILE:
+
+You are analyzing a specific file:
+- File: ${file}
+- Lines: 1-${endLine} (whole file)
+
+You MUST:
+- Analyze ONLY this specific file — do not search for or report issues in other files
+- You may read other files to understand context (e.g., imports, type definitions, data flow), but only report issues in this file
+- Do NOT scan the broader repository for other instances of this vulnerability pattern
+`;
+}
+/**
+ * Walk a directory tree recursively and yield every file path (POSIX-normalized,
+ * relative to `root`). Skips well-known noise directories so glob patterns do
+ * not have to opt out of `node_modules`/`.git` etc. on every check.
+ *
+ * Symlinks are NOT followed: `Dirent.isDirectory()` returns true only for real
+ * directories, so symlinked directories (and any cycles they could form) are
+ * skipped. Symlinked files are also skipped (`isFile()` returns false). This
+ * is the safe default — there's no general way to detect a hostile cycle, and
+ * source repos rarely rely on symlink-to-directory for layout.
+ */
+async function* walkFiles(root) {
+    async function* walk(dir) {
+        let entries;
+        try {
+            entries = await readdir(dir, { withFileTypes: true });
+        }
+        catch {
+            return;
+        }
+        for (const entry of entries) {
+            const full = join(dir, entry.name);
+            if (entry.isDirectory()) {
+                if (ALWAYS_SKIP_DIRS.has(entry.name))
+                    continue;
+                yield* walk(full);
+            }
+            else if (entry.isFile()) {
+                const rel = relative(root, full).split(sep).join('/');
+                yield rel;
+            }
+            // Symlinks (entry.isSymbolicLink()) are intentionally skipped — see
+            // the function-level comment above.
+        }
+    }
+    yield* walk(root);
+}
+/**
+ * Count the number of lines in a file by streaming and counting `\n` bytes.
+ * Constant memory; safe for very large files.
+ *
+ * Returns 1 for empty files so that `endLine >= startLine` always holds.
+ * Returns `undefined` if the file cannot be read or exceeds `MAX_FILE_SIZE_BYTES`
+ * — callers should treat that as "skip this target" and log accordingly.
+ */
+async function countLines(filePath) {
+    let size;
+    try {
+        const st = await stat(filePath);
+        size = st.size;
+    }
+    catch (err) {
+        logDebug(TAG, `failed to stat ${filePath} for line count: ${err instanceof Error ? err.message : String(err)}`);
+        return undefined;
+    }
+    if (size > MAX_FILE_SIZE_BYTES) {
+        logDebug(TAG, `skipping ${filePath}: size ${size} bytes exceeds limit ${MAX_FILE_SIZE_BYTES}`);
+        return undefined;
+    }
+    if (size === 0)
+        return 1;
+    return new Promise((resolvePromise) => {
+        let newlineCount = 0;
+        let endsWithNewline = false;
+        const stream = createReadStream(filePath);
+        stream.on('data', (chunk) => {
+            // Treat the stream as bytes — counting `\n` bytes is safe for UTF-8 because
+            // 0x0A never appears as a continuation byte of a multi-byte sequence.
+            const buf = typeof chunk === 'string' ? Buffer.from(chunk) : chunk;
+            for (let i = 0; i < buf.length; i++) {
+                if (buf[i] === 0x0a)
+                    newlineCount++;
+            }
+            if (buf.length > 0)
+                endsWithNewline = buf[buf.length - 1] === 0x0a;
+        });
+        stream.on('error', (err) => {
+            logDebug(TAG, `failed to read ${filePath} for line count: ${err.message}`);
+            resolvePromise(undefined);
+        });
+        stream.on('end', () => {
+            // If the file ends with a newline, the count of \n equals the number of lines.
+            // If it does not, the trailing partial line adds one to the count.
+            const lines = endsWithNewline ? newlineCount : newlineCount + 1;
+            resolvePromise(lines || 1);
+        });
+    });
+}
+/**
+ * Discover whole-file targets in `repoPath` matching `pattern`.
+ *
+ * Exposed as a named export (rather than only the registry-facing object) so
+ * unit tests can exercise the matching logic directly without going through
+ * the SecurityCheck plumbing.
+ */
+export async function discoverGlobTargets(pattern, repoPath, maxTargets) {
+    if (!pattern || pattern.trim() === '') {
+        throw new Error(formatError(ERROR_CODES.E2004, 'Glob discovery requires a non-empty "glob" pattern'));
+    }
+    const matcher = picomatch(pattern, { dot: false });
+    const matches = [];
+    for await (const rel of walkFiles(repoPath)) {
+        if (matcher(rel)) {
+            matches.push(rel);
+        }
+    }
+    // Sort BEFORE applying maxTargets so that the truncated subset is deterministic
+    // across platforms (readdir order varies by filesystem). Without this, a glob
+    // with maxTargets=N could analyze a different N files on Linux vs Windows vs
+    // macOS, making findings non-reproducible.
+    matches.sort();
+    const limited = typeof maxTargets === 'number' ? matches.slice(0, maxTargets) : matches;
+    // Compute line counts in parallel; drop unreadable / oversized files.
+    const lineCounts = await Promise.all(limited.map((rel) => countLines(join(repoPath, rel))));
+    const usable = [];
+    for (let i = 0; i < limited.length; i++) {
+        const endLine = lineCounts[i];
+        if (endLine === undefined)
+            continue; // unreadable / oversized — already logged
+        usable.push({ rel: limited[i], endLine });
+    }
+    // Surface a single summary log when files were dropped, so users debugging
+    // "why is my big bundle missing from the target set?" don't have to enable
+    // trace-level logging to find the per-file skip messages from countLines.
+    const skippedCount = limited.length - usable.length;
+    if (skippedCount > 0) {
+        logDebug(TAG, `${skippedCount} matched file${skippedCount === 1 ? '' : 's'} skipped (unreadable or > ${MAX_FILE_SIZE_BYTES} bytes)`);
+    }
+    return usable.map(({ rel, endLine }, idx) => ({
+        file: rel,
+        startLine: 1,
+        endLine,
+        label: `[file ${idx + 1}/${usable.length}]`,
+        promptEnrichment: buildTargetPromptEnrichment(rel, endLine),
+    }));
+}
+export const globDiscovery = {
+    name: 'glob',
+    defaultGenericPrompt: DEFAULT_GENERIC_PROMPT,
+    needsInstructions: true,
+    // Opted out deliberately. `supportsDiffFilter` became a required field on
+    // TargetDiscovery after this discovery was written (#227), and the diff
+    // filter is built around findings with meaningful line ranges. Glob targets
+    // are whole files, so filtering them has never been designed or tested for
+    // this shape. Opting out preserves the behaviour this discovery was built
+    // and tested with: a glob check scans every matching file. Enabling it later
+    // is a small change plus tests for the whole-file case.
+    supportsDiffFilter: false,
+    async discover(check, repoPath, _options) {
+        const checkTarget = check.checkTarget;
+        if (!checkTarget.glob) {
+            throw new Error(formatError(ERROR_CODES.E2004, `Check "${check.id}" uses glob discovery but has no "glob" pattern in its check definition`));
+        }
+        logDebug(TAG, `Running glob discovery for check "${check.id}": ${checkTarget.glob}`);
+        const targets = await discoverGlobTargets(checkTarget.glob, repoPath, checkTarget.maxTargets);
+        logDebug(TAG, `Discovered ${targets.length} files matching pattern`);
+        return targets;
+    },
+};
+//# sourceMappingURL=glob-discovery.js.map

package/dist/discoveries/glob-discovery.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"glob-discovery.js","sourceRoot":"","sources":["../../src/discoveries/glob-discovery.ts"],"names":[],"mappings":"AAAA;;;;;;;;;GASG;AAEH,OAAO,EAAE,OAAO,EAAE,IAAI,EAAE,MAAM,kBAAkB,CAAC;AACjD,OAAO,EAAE,gBAAgB,EAAE,MAAM,SAAS,CAAC;AAC3C,OAAO,EAAE,IAAI,EAAE,QAAQ,EAAE,GAAG,EAAE,MAAM,WAAW,CAAC;AAChD,OAAO,SAAS,MAAM,WAAW,CAAC;AAClC,OAAO,EAAE,QAAQ,EAAE,MAAM,eAAe,CAAC;AACzC,OAAO,EAAE,WAAW,EAAE,WAAW,EAAE,MAAM,mBAAmB,CAAC;AAC7D,OAAO,EAAE,sBAAsB,EAAE,MAAM,gBAAgB,CAAC;AAIxD,MAAM,GAAG,GAAG,gBAAgB,CAAC;AAE7B;;;;GAIG;AACH,MAAM,mBAAmB,GAAG,EAAE,GAAG,IAAI,GAAG,IAAI,CAAC,CAAC,SAAS;AAEvD,4EAA4E;AAC5E,MAAM,gBAAgB,GAAG,IAAI,GAAG,CAAC;IAC/B,MAAM;IACN,cAAc;IACd,OAAO;IACP,MAAM;IACN,aAAa;IACb,MAAM;IACN,aAAa;IACb,eAAe;IACf,MAAM;IACN,OAAO;IACP,OAAO;IACP,OAAO;IACP,QAAQ;IACR,OAAO;IACP,SAAS;CACV,CAAC,CAAC;AAEH,SAAS,2BAA2B,CAAC,IAAY,EAAE,OAAe;IAChE,OAAO;;;UAGC,IAAI;aACD,OAAO;;;;;;CAMnB,CAAC;AACF,CAAC;AAED;;;;;;;;;;GAUG;AACH,KAAK,SAAS,CAAC,CAAC,SAAS,CAAC,IAAY;IACpC,KAAK,SAAS,CAAC,CAAC,IAAI,CAAC,GAAW;QAC9B,IAAI,OAAmC,CAAC;QACxC,IAAI,CAAC;YACH,OAAO,GAAG,MAAM,OAAO,CAAC,GAAG,EAAE,EAAE,aAAa,EAAE,IAAI,EAAE,CAAC,CAAC;QACxD,CAAC;QAAC,MAAM,CAAC;YACP,OAAO;QACT,CAAC;QACD,KAAK,MAAM,KAAK,IAAI,OAAO,EAAE,CAAC;YAC5B,MAAM,IAAI,GAAG,IAAI,CAAC,GAAG,EAAE,KAAK,CAAC,IAAI,CAAC,CAAC;YACnC,IAAI,KAAK,CAAC,WAAW,EAAE,EAAE,CAAC;gBACxB,IAAI,gBAAgB,CAAC,GAAG,CAAC,KAAK,CAAC,IAAI,CAAC;oBAAE,SAAS;gBAC/C,KAAK,CAAC,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC;YACpB,CAAC;iBAAM,IAAI,KAAK,CAAC,MAAM,EAAE,EAAE,CAAC;gBAC1B,MAAM,GAAG,GAAG,QAAQ,CAAC,IAAI,EAAE,IAAI,CAAC,CAAC,KAAK,CAAC,GAAG,CAAC,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC;gBACtD,MAAM,GAAG,CAAC;YACZ,CAAC;YACD,oEAAoE;YACpE,oCAAoC;QACtC,CAAC;IACH,CAAC;IACD,KAAK,CAAC,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC;AACpB,CAAC;AAED;;;;;;;GAOG;AACH,KAAK,UAAU,UAAU,CAAC,QAAgB;IACxC,IAAI,IAAY,CAAC;IACjB,IAAI,CAAC;QACH,MAAM,EAAE,GAAG,MAAM,IAAI,CAAC,QAAQ,CAAC,CAAC;QAChC,IAAI,GAAG,EAAE,CAAC,IAAI,CAAC;IACjB,CAAC;IAAC,OAAO,GAAG,EAAE,CAAC;QACb,QAAQ,CAAC,GAAG,EAAE,kBAAkB,QAAQ,oBAAoB,GAAG,YAAY,KAAK,CAAC,CAAC,CAAC,GAAG,CAAC,OAAO,CAAC,CAAC,CAAC,MAAM,CAAC,GAAG,CAAC,EAAE,CAAC,CAAC;QAChH,OAAO,SAAS,CAAC;IACnB,CAAC;IAED,IAAI,IAAI,GAAG,mBAAmB,EAAE,CAAC;QAC/B,QAAQ,CAAC,GAAG,EAAE,YAAY,QAAQ,UAAU,IAAI,wBAAwB,mBAAmB,EAAE,CAAC,CAAC;QAC/F,OAAO,SAAS,CAAC;IACnB,CAAC;IAED,IAAI,IAAI,KAAK,CAAC;QAAE,OAAO,CAAC,CAAC;IAEzB,OAAO,IAAI,OAAO,CAAqB,CAAC,cAAc,EAAE,EAAE;QACxD,IAAI,YAAY,GAAG,CAAC,CAAC;QACrB,IAAI,eAAe,GAAG,KAAK,CAAC;QAC5B,MAAM,MAAM,GAAG,gBAAgB,CAAC,QAAQ,CAAC,CAAC;QAC1C,MAAM,CAAC,EAAE,CAAC,MAAM,EAAE,CAAC,KAAsB,EAAE,EAAE;YAC3C,4EAA4E;YAC5E,sEAAsE;YACtE,MAAM,GAAG,GAAG,OAAO,KAAK,KAAK,QAAQ,CAAC,CAAC,CAAC,MAAM,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC,KAAK,CAAC;YACnE,KAAK,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,GAAG,CAAC,MAAM,EAAE,CAAC,EAAE,EAAE,CAAC;gBACpC,IAAI,GAAG,CAAC,CAAC,CAAC,KAAK,IAAI;oBAAE,YAAY,EAAE,CAAC;YACtC,CAAC;YACD,IAAI,GAAG,CAAC,MAAM,GAAG,CAAC;gBAAE,eAAe,GAAG,GAAG,CAAC,GAAG,CAAC,MAAM,GAAG,CAAC,CAAC,KAAK,IAAI,CAAC;QACrE,CAAC,CAAC,CAAC;QACH,MAAM,CAAC,EAAE,CAAC,OAAO,EAAE,CAAC,GAAG,EAAE,EAAE;YACzB,QAAQ,CAAC,GAAG,EAAE,kBAAkB,QAAQ,oBAAoB,GAAG,CAAC,OAAO,EAAE,CAAC,CAAC;YAC3E,cAAc,CAAC,SAAS,CAAC,CAAC;QAC5B,CAAC,CAAC,CAAC;QACH,MAAM,CAAC,EAAE,CAAC,KAAK,EAAE,GAAG,EAAE;YACpB,+EAA+E;YAC/E,mEAAmE;YACnE,MAAM,KAAK,GAAG,eAAe,CAAC,CAAC,CAAC,YAAY,CAAC,CAAC,CAAC,YAAY,GAAG,CAAC,CAAC;YAChE,cAAc,CAAC,KAAK,IAAI,CAAC,CAAC,CAAC;QAC7B,CAAC,CAAC,CAAC;IACL,CAAC,CAAC,CAAC;AACL,CAAC;AAED;;;;;;GAMG;AACH,MAAM,CAAC,KAAK,UAAU,mBAAmB,CACvC,OAAe,EACf,QAAgB,EAChB,UAAmB;IAEnB,IAAI,CAAC,OAAO,IAAI,OAAO,CAAC,IAAI,EAAE,KAAK,EAAE,EAAE,CAAC;QACtC,MAAM,IAAI,KAAK,CACb,WAAW,CAAC,WAAW,CAAC,KAAK,EAAE,oDAAoD,CAAC,CACrF,CAAC;IACJ,CAAC;IAED,MAAM,OAAO,GAAG,SAAS,CAAC,OAAO,EAAE,EAAE,GAAG,EAAE,KAAK,EAAE,CAAC,CAAC;IACnD,MAAM,OAAO,GAAa,EAAE,CAAC;IAE7B,IAAI,KAAK,EAAE,MAAM,GAAG,IAAI,SAAS,CAAC,QAAQ,CAAC,EAAE,CAAC;QAC5C,IAAI,OAAO,CAAC,GAAG,CAAC,EAAE,CAAC;YACjB,OAAO,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC;QACpB,CAAC;IACH,CAAC;IAED,gFAAgF;IAChF,8EAA8E;IAC9E,6EAA6E;IAC7E,2CAA2C;IAC3C,OAAO,CAAC,IAAI,EAAE,CAAC;IACf,MAAM,OAAO,GAAG,OAAO,UAAU,KAAK,QAAQ,CAAC,CAAC,CAAC,OAAO,CAAC,KAAK,CAAC,CAAC,EAAE,UAAU,CAAC,CAAC,CAAC,CAAC,OAAO,CAAC;IAExF,sEAAsE;IACtE,MAAM,UAAU,GAAG,MAAM,OAAO,CAAC,GAAG,CAClC,OAAO,CAAC,GAAG,CAAC,CAAC,GAAG,EAAE,EAAE,CAAC,UAAU,CAAC,IAAI,CAAC,QAAQ,EAAE,GAAG,CAAC,CAAC,CAAC,CACtD,CAAC;IACF,MAAM,MAAM,GAAuC,EAAE,CAAC;IACtD,KAAK,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,OAAO,CAAC,MAAM,EAAE,CAAC,EAAE,EAAE,CAAC;QACxC,MAAM,OAAO,GAAG,UAAU,CAAC,CAAC,CAAC,CAAC;QAC9B,IAAI,OAAO,KAAK,SAAS;YAAE,SAAS,CAAC,0CAA0C;QAC/E,MAAM,CAAC,IAAI,CAAC,EAAE,GAAG,EAAE,OAAO,CAAC,CAAC,CAAC,EAAE,OAAO,EAAE,CAAC,CAAC;IAC5C,CAAC;IAED,2EAA2E;IAC3E,2EAA2E;IAC3E,0EAA0E;IAC1E,MAAM,YAAY,GAAG,OAAO,CAAC,MAAM,GAAG,MAAM,CAAC,MAAM,CAAC;IACpD,IAAI,YAAY,GAAG,CAAC,EAAE,CAAC;QACrB,QAAQ,CACN,GAAG,EACH,GAAG,YAAY,gBAAgB,YAAY,KAAK,CAAC,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,GAAG,6BAA6B,mBAAmB,SAAS,CACtH,CAAC;IACJ,CAAC;IAED,OAAO,MAAM,CAAC,GAAG,CAAC,CAAC,EAAE,GAAG,EAAE,OAAO,EAAE,EAAE,GAAG,EAAE,EAAE,CAAC,CAAC;QAC5C,IAAI,EAAE,GAAG;QACT,SAAS,EAAE,CAAC;QACZ,OAAO;QACP,KAAK,EAAE,SAAS,GAAG,GAAG,CAAC,IAAI,MAAM,CAAC,MAAM,GAAG;QAC3C,gBAAgB,EAAE,2BAA2B,CAAC,GAAG,EAAE,OAAO,CAAC;KAC5D,CAAC,CAAC,CAAC;AACN,CAAC;AAED,MAAM,CAAC,MAAM,aAAa,GAAoB;IAC5C,IAAI,EAAE,MAAM;IACZ,oBAAoB,EAAE,sBAAsB;IAC5C,iBAAiB,EAAE,IAAI;IACvB,0EAA0E;IAC1E,wEAAwE;IACxE,4EAA4E;IAC5E,2EAA2E;IAC3E,0EAA0E;IAC1E,6EAA6E;IAC7E,wDAAwD;IACxD,kBAAkB,EAAE,KAAK;IAEzB,KAAK,CAAC,QAAQ,CACZ,KAAoB,EACpB,QAAgB,EAChB,QAA2B;QAE3B,MAAM,WAAW,GAAG,KAAK,CAAC,WAAY,CAAC;QAEvC,IAAI,CAAC,WAAW,CAAC,IAAI,EAAE,CAAC;YACtB,MAAM,IAAI,KAAK,CACb,WAAW,CACT,WAAW,CAAC,KAAK,EACjB,UAAU,KAAK,CAAC,EAAE,yEAAyE,CAC5F,CACF,CAAC;QACJ,CAAC;QAED,QAAQ,CAAC,GAAG,EAAE,qCAAqC,KAAK,CAAC,EAAE,MAAM,WAAW,CAAC,IAAI,EAAE,CAAC,CAAC;QAErF,MAAM,OAAO,GAAG,MAAM,mBAAmB,CAAC,WAAW,CAAC,IAAI,EAAE,QAAQ,EAAE,WAAW,CAAC,UAAU,CAAC,CAAC;QAE9F,QAAQ,CAAC,GAAG,EAAE,cAAc,OAAO,CAAC,MAAM,yBAAyB,CAAC,CAAC;QACrE,OAAO,OAAO,CAAC;IACjB,CAAC;CACF,CAAC"}

package/dist/discoveries/script-discovery.d.ts

@@ -0,0 +1,117 @@
+/**
+ * Script-based target discovery.
+ *
+ * Runs a user-provided script (`node` or `bash`) in the target repository
+ * and parses its stdout to produce discovery targets. Useful when target
+ * discovery is too custom for Semgrep / SARIF / OpenAnt — e.g. parsing an
+ * OpenAPI spec, querying a build manifest, walking a database schema, etc.
+ *
+ * ─── Trust model ─────────────────────────────────────────────────────────────
+ *
+ * Scripts run with the full privileges of the aghast process. They are not
+ * sandboxed. Treat any script referenced by a check definition as trusted
+ * code — only enable script-discovery checks whose script you (or your team)
+ * have read and audited.
+ *
+ * Defensive measures applied here:
+ * - Script path must resolve INSIDE the configured check folder or repo
+ *   (no `..` escape, no absolute paths to /etc/shadow, etc.).
+ * - Symlink-aware: both the script path and `cwd` are realpath'd and
+ *   re-checked against the repo root, so a symlink inside the check folder
+ *   pointing outside the repo cannot be used to execute external code.
+ * - Spawned with `shell: false` and array args — no shell interpolation, no
+ *   command injection from check-definition fields.
+ * - Spawned with a curated environment: only a small allow-list of neutral
+ *   env vars (PATH, HOME, locale, temp dirs, Windows essentials) is forwarded.
+ *   As defence-in-depth, any var matching a known-secret pattern (`*_KEY`,
+ *   `*_TOKEN`, `*_SECRET`, `*_PASSWORD`, or `ANTHROPIC_*` / `OPENAI_*` /
+ *   `GITHUB_*` / `AGHAST_*` / `AWS_*` / `GCP_*` / `AZURE_*` prefixes) is
+ *   also explicitly dropped. Allow-list matching is case-insensitive so
+ *   the Windows-style `Path` / `SystemRoot` casing works the same as POSIX
+ *   `PATH`.
+ * - cwd, when supplied, must resolve (and realpath) inside the repo root.
+ * - Hard timeout (default 30s, configurable per check).
+ * - Stdout is bounded (8 MiB) to prevent memory exhaustion from runaway scripts.
+ * - Both the line-format and JSON-format output paths cap the number of
+ *   parsed targets (MAX_LINE_TARGETS / MAX_JSON_TARGETS) to bound memory
+ *   even if a script floods within the stdout cap.
+ * - Output is parsed as data (lines / JSON) — never `eval`-ed.
+ * - File paths returned by the script are validated to be relative & inside
+ *   the repo (no `..` escape, no absolute paths) before they are forwarded
+ *   to the AI prompt or snippet extractor.
+ * - On `bash` scripts, Windows is unsupported (no system bash) and a clear
+ *   error is raised rather than silently falling back.
+ */
+import type { TargetDiscovery } from '../discovery.js';
+/**
+ * Build a curated environment for the spawned discovery script.
+ *
+ * Discovery scripts need to walk the repo, possibly run small helper tools,
+ * but they never need API credentials. We start from a minimal allow-list of
+ * "neutral" env vars (PATH/locale/temp dirs) and then ALSO drop any var whose
+ * name matches a known-secret pattern even if it slipped through. Both layers
+ * are defence-in-depth.
+ *
+ * Exported for testing.
+ */
+export declare function buildScriptEnv(parent?: NodeJS.ProcessEnv): NodeJS.ProcessEnv;
+interface ScriptTargetInput {
+    file: string;
+    startLine?: number;
+    endLine?: number;
+    message?: string;
+    snippet?: string;
+}
+/**
+ * True when `candidateAbs` lives inside `rootAbs`. Both must be absolute.
+ *
+ * Two ways to be outside, and both matter:
+ *   - a `..` prefix, the ordinary case;
+ *   - an absolute result, which is what `relative()` returns on Windows when
+ *     the two paths are on different drives (`C:\repo` vs `D:\checks`).
+ *
+ * Exported so the rule can be unit-tested against Windows-style paths on any
+ * platform. It exists as a single function because it previously did not: two
+ * call sites implemented the rule separately, one omitted the absolute-path
+ * case, and cross-drive layouts were rejected as a result.
+ *
+ * Pass a `pathApi` of `path.win32` or `path.posix` to test either flavour.
+ */
+export declare function isInsideRoot(rootAbs: string, candidateAbs: string, pathApi?: {
+    relative: (a: string, b: string) => string;
+    isAbsolute: (p: string) => boolean;
+}): boolean;
+/**
+ * Parse stdout into a list of discovery target inputs.
+ * - `lines`: one file path per non-empty, non-comment line.
+ * - `json-array`: a JSON array of strings (file paths) or objects.
+ * - `json-object`: an object with a `targets` array of strings or objects.
+ *
+ * Both `lines` and JSON formats cap parsed targets at MAX_LINE_TARGETS /
+ * MAX_JSON_TARGETS (100k each); the rest are dropped with a warning log.
+ *
+ * Output is treated as DATA: never eval'd, never interpolated into commands.
+ */
+export declare function parseScriptOutput(raw: string, format: 'lines' | 'json-array' | 'json-object'): ScriptTargetInput[];
+interface RunScriptResult {
+    stdout: string;
+    exitCode: number | null;
+    signal: NodeJS.Signals | null;
+    stderr: string;
+}
+/**
+ * Execute the script as a child process and capture stdout.
+ * - Uses `spawn` with `shell: false`: never invoke a shell, never interpolate.
+ * - Passes a curated env (no API keys / tokens) — see `buildScriptEnv`.
+ * - Bounds stdout/stderr buffers and kills the process on timeout.
+ */
+export declare function runScript(opts: {
+    command: string;
+    args: string[];
+    cwd: string;
+    timeoutMs: number;
+    env?: NodeJS.ProcessEnv;
+}): Promise<RunScriptResult>;
+export declare const scriptDiscovery: TargetDiscovery;
+export {};
+//# sourceMappingURL=script-discovery.d.ts.map

package/dist/discoveries/script-discovery.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"script-discovery.d.ts","sourceRoot":"","sources":["../../src/discoveries/script-discovery.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GA2CG;AAQH,OAAO,KAAK,EAAE,eAAe,EAAsC,MAAM,iBAAiB,CAAC;AAoB3F;;;;;;;;;;GAUG;AACH,wBAAgB,cAAc,CAAC,MAAM,GAAE,MAAM,CAAC,UAAwB,GAAG,MAAM,CAAC,UAAU,CA8DzF;AAED,UAAU,iBAAiB;IACzB,IAAI,EAAE,MAAM,CAAC;IACb,SAAS,CAAC,EAAE,MAAM,CAAC;IACnB,OAAO,CAAC,EAAE,MAAM,CAAC;IACjB,OAAO,CAAC,EAAE,MAAM,CAAC;IACjB,OAAO,CAAC,EAAE,MAAM,CAAC;CAClB;AAED;;;;;;;;;;;;;;GAcG;AACH,wBAAgB,YAAY,CAC1B,OAAO,EAAE,MAAM,EACf,YAAY,EAAE,MAAM,EACpB,OAAO,GAAE;IAAE,QAAQ,EAAE,CAAC,CAAC,EAAE,MAAM,EAAE,CAAC,EAAE,MAAM,KAAK,MAAM,CAAC;IAAC,UAAU,EAAE,CAAC,CAAC,EAAE,MAAM,KAAK,OAAO,CAAA;CAC/D,GACzB,OAAO,CAGT;AA4GD;;;;;;;;;;GAUG;AACH,wBAAgB,iBAAiB,CAC/B,GAAG,EAAE,MAAM,EACX,MAAM,EAAE,OAAO,GAAG,YAAY,GAAG,aAAa,GAC7C,iBAAiB,EAAE,CAsIrB;AAED,UAAU,eAAe;IACvB,MAAM,EAAE,MAAM,CAAC;IACf,QAAQ,EAAE,MAAM,GAAG,IAAI,CAAC;IACxB,MAAM,EAAE,MAAM,CAAC,OAAO,GAAG,IAAI,CAAC;IAC9B,MAAM,EAAE,MAAM,CAAC;CAChB;AAED;;;;;GAKG;AACH,wBAAgB,SAAS,CAAC,IAAI,EAAE;IAC9B,OAAO,EAAE,MAAM,CAAC;IAChB,IAAI,EAAE,MAAM,EAAE,CAAC;IACf,GAAG,EAAE,MAAM,CAAC;IACZ,SAAS,EAAE,MAAM,CAAC;IAClB,GAAG,CAAC,EAAE,MAAM,CAAC,UAAU,CAAC;CACzB,GAAG,OAAO,CAAC,eAAe,CAAC,CAsF3B;AAgCD,eAAO,MAAM,eAAe,EAAE,eAyM7B,CAAC"}