User-facing surface for Ruflo's self-learning system: 6 neural_* + 10 hooks_intelligence_* + 6 routing/meta hooks + 3 hooks_model-* + 4 SONA/MicroLoRA tools…

User-facing surface for Ruflo's self-learning system. Wraps 29 intelligence-related MCP tools across four families into discoverable skills, commands, and the canonical 4-step pipeline (RETRIEVE → JUDGE → DISTILL → CONSOLIDATE). Coordinates with ruflo-agentdb (namespace convention), ruflo-ruvector (trajectory recording substrate), and ruflo-browser (consumes trajectory hooks for session replay).
Status: ADR-0001 implemented. Plugin v0.3.0 targets
@claude-flow/cliv3.6.x.
/plugin marketplace add ruvnet/ruflo
/plugin install ruflo-intelligence@ruflo
@claude-flow/cli v3.6 major+minor.bash plugins/ruflo-intelligence/scripts/smoke.sh is the contract.| Family | Count | Source |
|---|---|---|
neural_* | 6 | v3/@claude-flow/cli/src/mcp-tools/neural-tools.ts:195, 312, 413, 539, 651, 706 |
hooks_intelligence_* (incl. dispatcher + reset) | 10 | v3/@claude-flow/cli/src/mcp-tools/hooks-tools.ts:2093, 2226, 2296, 2355, 2404, 2556, 2634, 2741, 2952, 3027 |
Routing & meta hooks (hooks_route, hooks_explain, hooks_pretrain, hooks_build-agents, hooks_metrics, hooks_transfer) | 6 | hooks-tools.ts:884, 1062, 1420, 1499, 1593, 1664 |
hooks_model-* (3-tier routing) | 3 | hooks-tools.ts:3797, 3844, 3879 |
ruvllm_sona_* + ruvllm_microlora_* | 4 | v3/@claude-flow/cli/src/mcp-tools/ruvllm-tools.ts:142, 169, 192, 222 |
| Total | 29 | — |
CLAUDE.md describes the V3 intelligence loop as four discrete phases. This plugin operationalizes them:
| Step | What happens | Tools |
|---|---|---|
| RETRIEVE | Pull relevant patterns + past trajectories from HNSW index | hooks_intelligence_pattern-search, agentdb_pattern-search, agentdb_semantic-route |
| JUDGE | Score retrieved candidates with verdicts (success / failure / partial) | hooks_intelligence_attention, neural_predict, hooks_explain |
| DISTILL | Extract the key learnings via LoRA / SONA adaptation | ruvllm_sona_adapt, ruvllm_microlora_adapt, neural_train, hooks_intelligence_learn |
| CONSOLIDATE | Prevent catastrophic forgetting via EWC++ | agentdb_consolidate, ruvllm_microlora_adapt --consolidate, neural_compress |
For an end-to-end run:
hooks_pretrain
→ hooks_intelligence_trajectory-start
→ (each step) hooks_intelligence_trajectory-step
→ hooks_intelligence_trajectory-end
→ hooks_intelligence_learn
→ ruvllm_sona_adapt # DISTILL
→ agentdb_consolidate # CONSOLIDATE
→ neural_compress # storage efficiency
hooks_transfer is the substrate plugin's most underused capability. It publishes learned patterns to IPFS (via Pinata) so a different project — or a different machine — can fetch and apply them. Use the intelligence-transfer skill or call directly:
# Publish patterns from this project to IPFS
mcp tool call hooks_transfer --json -- '{"action": "store", "patterns": [...]}'
# Fetch and apply patterns from a CID
mcp tool call hooks_transfer --json -- '{"action": "load", "cid": "QmXyz..."}'
# Mirror an entire project's patterns
mcp tool call hooks_transfer --json -- '{"action": "from-project", "source": "/path/to/project"}'
Prerequisite: PINATA_API_JWT (or the equivalent endpoint env vars) must be configured. Without it, hooks_transfer returns a structured success: false with the missing-config error.
Several Claude Code hooks fire intelligence-side writes:
| Hook | Tool invoked | Target |
|---|---|---|
pre-task | hooks_route + hooks_intelligence_pattern-search | RETRIEVE phase |
post-task --train-neural | agentdb_pattern-store (ReasoningBank) → falls back to memory_store --namespace pattern | DISTILL phase, writes to pattern namespace |
pretrain (one-shot) | hooks_pretrain → seeds memory_store --namespace patterns | Bootstrap, writes to patterns namespace (plural) |
| Trajectory hooks (ruvector substrate) | intelligence_trajectory-* | Recorded by ruflo-ruvector; consumed by this plugin's pattern-store |
Pluralization gotcha: ReasoningBank fallback writes to
pattern(singular). Thepretrainhook writes topatterns(plural). They are different namespaces. Seeruflo-agentdbADR-0001 §"Namespace convention" for the canonical contract.
This plugin defers to ruflo-agentdb ADR-0001 for namespace conventions. Three reserved namespaces are read by the intelligence pipeline:
| Namespace | Read by | Source |
|---|---|---|
pattern | hooks_intelligence_pattern-search, agentdb_pattern-search | ReasoningBank fallback target |
patterns (plural) | hooks_pretrain, neural_train corpus | distinct from pattern |
claude-memories | memory_search_unified (default include) | Claude Code auto-memory bridge |
Do not invent new top-level namespaces for intelligence purposes — the convention is owned upstream.
The plugin claims EWC++ consolidation; here's how to actually invoke it:
hooks_intelligence_learn to register the outcome.agentdb_consolidate to fold patterns into the long-term store under EWC++ semantics.ruvllm_microlora_adapt with the --consolidate flag to apply Elastic Weight Consolidation on the adapter's weight deltas. This prevents catastrophic forgetting when the adapter is trained on a new domain.Without these calls, fresh trajectories overwrite older patterns without protection — the system "forgets". The pipeline diagram above bakes consolidation into step 4 deliberately.
hooks_intelligence accepts a mode parameter that selects the active learning architecture:
| Mode | When to use |
|---|---|
balanced (default) | General-purpose: SONA + HNSW retrieval, no MoE specialization |
sona | Single-domain specialization with SONA adaptation |
moe | Multi-domain expert routing — recommended when tasks span 3+ distinct domains |
hnsw | Pure pattern retrieval, no online adaptation |
Configure once via mcp tool call hooks_intelligence -- '{"mode": "moe", "enableSona": true}' and let the dispatcher route subsequent learning calls.
/intelligence — Dashboard: stats, metrics, model-tier distribution, routing rationale on demand/neural — Neural training and prediction (train, status, patterns, predict, optimize, compress)neural-train — Train SONA + MicroLoRA patterns from successful tasksintelligence-route — Route tasks using learned patterns; produces a hooks_explain rationaleintelligence-transfer — Publish/fetch patterns via IPFS (hooks_transfer)A function-hook mod ships beside the skills (ADR-445 pattern). Needs a Claude Code with mods (2.1.287+); older builds ignore it. No network, no process spawning: it only tightens calls to this plugin's own tools and reads through tools already connected.
| Piece | Default | What it does |
|---|---|---|
| Write guard | on | Refuses a secret in a learned pattern, trajectory step/end, neural_train or hooks_transfer call, or a memory_store into a pattern namespace. A transfer also refuses an email address (IPFS is public). |
| Reset confirm | on | Refuses hooks_intelligence-reset unless the call carries confirm: true. |
/intelligence-mod | — | status, scan <text>, stats (calls hooks_intelligence_stats through the connected tool); answered locally, no model call. |
| Status file | — | .claude-flow/intelligence-mod/status.json (version, updatedMs, mode flags and counters); written at session start and when a counter changes. |
Options (userConfig): guard on\|off, confirmReset on\|off. Refusals never echo the value they matched.
claude plugin test plugins/ruflo-intelligence # 11 tests
ruflo-agentdb — substrate for HNSW + namespace contract; agentdb_pattern-* is this plugin's storage backendruflo-ruvector — trajectory hooks substrate; intelligence_trajectory-* calls land in ruvector's persisted trajectoriesruflo-browser — consumes trajectory hooks for session replay (ADR-0001 there)ruflo-daa — Dynamic Agentic Architecture; cognitive patterns feed routing as inputsMIT
hooks/register.ts 76 lines1import type { Hook, Register } from 'claude-code'
2
3import { answer } from './command'
4import { verdict } from './guard'
5import { modeOf, readOptions, type ModOptions } from './options'
6import { newStats, STATUS_PATH, statusText, type Stats } from './status'
7import { findTool } from './tools'
8
9const TEXT_CAP = 20_000
10
11type Dollar = Parameters<Hook<'session.start'>>[0]
12
13/** Everything one session of the mod keeps: its settings, its counters and the project root. */
14type Session = { readonly opts: ModOptions; readonly stats: Stats; root?: string }
15
16async function flush($: Dollar, s: Session): Promise<void> {
17 if (s.root === undefined) return
18 try {
19 await $.fs.write(`${s.root}/${STATUS_PATH}`, statusText(s.stats, modeOf(s.opts), await $.clock.now()))
20 } catch {
21 /* the status file is a courtesy */
22 }
23}
24
25/** A read-only call through an already-connected tool: its first text block, or undefined when absent or errored. */
26async function peek($: Dollar, suffix: string, args: Record<string, unknown>): Promise<string | undefined> {
27 const hit = findTool(await $.tool.list(), suffix)
28 if (!hit) return undefined
29 const res = await $.mcp.call(hit.server, hit.tool, args)
30 return res.isError ? undefined : (res.content.find(b => b.type === 'text')?.text ?? '').slice(0, TEXT_CAP)
31}
32
33/**
34 * Intelligence as a mod (ADR-445 pattern): a tighten-only guard on this plugin's own tools, `/intelligence-mod`, and a status file the console
35 * reads. No network, no process spawning: only tools already connected.
36 */
37export const register: Register = (on, options) => {
38 const s: Session = { opts: readOptions(options), stats: newStats() }
39
40 on('session.start', async ($, e, next) => {
41 const result = await next(e)
42 s.root = (await $.session.root()) as string | undefined
43 try {
44 await $.command.register({ name: 'intelligence-mod', description: 'Intelligence mod: status, scan <text>, stats' })
45 } catch {
46 /* a name taken by another plugin must not stop the mod */
47 }
48 await flush($, s)
49 return result
50 })
51
52 if (s.opts.guard) {
53 on('tool.call', async ($, e, next) => {
54 const reason = verdict(e.tool, e, s.opts)
55 if (reason === undefined) return next(e)
56 s.stats.blocked++
57 s.stats.lastDenied = e.tool
58 await flush($, s)
59 return { deny: reason }
60 })
61 }
62
63 /** `/intelligence-mod` is answered locally and takes no model turn. */
64 on('command.run', { command: 'intelligence-mod' }, async ($, e) => {
65 s.stats.commands++
66 const text = await answer(typeof e.args === 'string' ? e.args : '', {
67 opts: s.opts,
68 stats: s.stats,
69 peek: (suffix, args) => peek($, suffix, args),
70 toolNames: async () => (await $.tool.list()).map(t => t.name),
71 })
72 await flush($, s)
73 return { text }
74 })
75}
76hooks/command.ts 41 lines1import { scan, tidy } from './screen'
2import type { ModOptions } from './options'
3import type { Stats } from './status'
4
5export type CommandDeps = {
6 readonly opts: ModOptions
7 readonly stats: Stats
8 readonly peek: (suffix: string, args: Record<string, unknown>) => Promise<string | undefined>
9 readonly toolNames: () => Promise<readonly string[]>
10}
11
12const HELP = ['/intelligence-mod status', '/intelligence-mod scan <text>', '/intelligence-mod stats'].join('\n')
13
14/** `/intelligence-mod` is answered locally and takes no model turn (the plugin's `/intelligence` and `/neural` are prompt commands). */
15export async function answer(args: string, deps: CommandDeps): Promise<string> {
16 const [verb = '', ...rest] = args.trim().split(/\s+/)
17 const arg = rest.join(' ')
18 const { opts, stats } = deps
19
20 if (verb === '' || verb === 'help') return HELP
21
22 if (verb === 'status') {
23 return `guard ${opts.guard ? 'on' : 'off'} · reset needs confirm ${opts.confirmReset ? 'on' : 'off'}\ncalls blocked ${stats.blocked} · scans ${stats.scans} · commands ${stats.commands}${stats.lastDenied ? ` · last denied ${stats.lastDenied}` : ''}`
24 }
25
26 if (verb === 'scan') {
27 if (arg === '') return 'usage: /intelligence-mod scan <text>'
28 stats.scans++
29 const found = scan(arg).secrets
30 return found.length ? `The guard would refuse a learning write of that (secrets: ${found.join(', ')}).` : 'No secret shape found: the guard would let that be learned.'
31 }
32
33 if (verb === 'stats') {
34 const text = await deps.peek('hooks_intelligence_stats', {})
35 if (text === undefined) return 'No intelligence tool is connected (or it errored). Connect the ruflo MCP server and try again.'
36 return text.trim() === '' ? 'No intelligence stats reported.' : tidy(text, 1200)
37 }
38
39 return `Unknown: ${verb}\n${HELP}`
40}
41hooks/guard.ts 34 lines1import { hasSecret, textsOf } from './screen'
2import { bare, namespaceOf } from './tools'
3import type { ModOptions } from './options'
4
5/** Learning writes: a secret in a stored pattern, trajectory or training row would be learned, recalled and, via transfer, published. */
6const LEARNERS = new Set(['hooks_intelligence_pattern-store', 'hooks_intelligence_trajectory-step', 'hooks_intelligence_trajectory-end', 'neural_train', 'hooks_transfer', 'memory_store'])
7const PATTERN_NS = /^(?:patterns?|intelligence|sona|neural|trajector)/i
8
9// `hooks_transfer` publishes to IPFS, which is public; an email address is personal data the plugin's own skill says to strip first.
10const EMAIL = /[A-Za-z0-9._%+-]{1,64}@[A-Za-z0-9-]{1,63}(?:\.[A-Za-z0-9-]{1,63})*\.[A-Za-z]{2,}/
11
12const isReset = (name: string) => name.replace(/_/g, '-') === 'hooks-intelligence-reset'
13
14/** The reason an intelligence call is refused, or undefined when it may go. Never names or echoes the value. */
15export function verdict(tool: string, input: unknown, opts: ModOptions): string | undefined {
16 const name = bare(tool)
17
18 if (isReset(name)) {
19 if (!opts.confirmReset || (input as { confirm?: unknown } | null)?.confirm === true) return undefined
20 return 'ruflo-intelligence: resetting the intelligence store erases every learned pattern. Ask the user first; once they agree, call again with confirm: true.'
21 }
22
23 if (!LEARNERS.has(name)) return undefined
24 if (name === 'memory_store' && !PATTERN_NS.test(namespaceOf(input))) return undefined
25 const texts = textsOf(input)
26 if (texts.some(hasSecret)) {
27 return 'ruflo-intelligence: this learning write holds what looks like a secret (a key, token or password). Learned patterns are recalled and can be published, so leave it out.'
28 }
29 if (name === 'hooks_transfer' && texts.some(t => EMAIL.test(t))) {
30 return 'ruflo-intelligence: this transfer holds an email address, and IPFS is public. Strip personal data first (aidefence_has_pii), then publish.'
31 }
32 return undefined
33}
34hooks/options.ts 17 lines1import type { PluginOptions } from 'claude-code'
2
3/** The plugin's `userConfig`, validated: a bad value is the default (both on). */
4export type ModOptions = { readonly guard: boolean; readonly confirmReset: boolean }
5
6// BEGIN SHARED FLAG (generated by scripts/sync-mod-screen.mjs; edit plugins/ruflo-agentdb/hooks/options.ts)
7export const flag = (value: unknown, fallback: boolean) =>
8 value === true || value === 'true' || value === 'on' ? true : value === false || value === 'false' || value === 'off' ? false : fallback
9// END SHARED FLAG
10
11export function readOptions(options: PluginOptions | undefined): ModOptions {
12 const o = options ?? {}
13 return { guard: flag(o.guard, true), confirmReset: flag(o.confirmReset, true) }
14}
15
16export const modeOf = (o: ModOptions): Record<string, boolean> => ({ guard: o.guard, confirmReset: o.confirmReset })
17hooks/status.ts 17 lines1/** Counters the mod keeps for the session and writes to `.claude-flow/intelligence-mod/status.json` for the console. */
2export type Stats = {
3 blocked: number
4 scans: number
5 commands: number
6 lastDenied?: string
7}
8
9export const newStats = (): Stats => ({ blocked: 0, scans: 0, commands: 0 })
10
11export const STATUS_PATH = '.claude-flow/intelligence-mod/status.json'
12
13/** The file's text: mode flags plus the counters; `version` lets the console refuse a shape it does not know. */
14export function statusText(stats: Stats, mode: Record<string, boolean>, nowMs: number): string {
15 return `${JSON.stringify({ version: 1, updatedMs: nowMs, ...mode, ...stats }, null, 2)}\n`
16}
17hooks/tools.ts 24 lines1import type { ToolInfo } from 'claude-code'
2
3/** `mcp__<server>__<tool>` into its two halves; tool names never hold a double underscore. */
4export function splitName(name: string): { server: string; tool: string } | undefined {
5 if (!name.startsWith('mcp__')) return undefined
6 const at = name.lastIndexOf('__')
7 return at > 5 ? { server: name.slice(5, at), tool: name.slice(at + 2) } : undefined
8}
9
10/** The bare tool name: `mcp__srv__memory_store` and `memory_store` both give `memory_store`. */
11export const bare = (name: string) => splitName(name)?.tool ?? name
12
13/** A connected MCP tool whose bare name is `suffix`, split into what `$.mcp.call` takes. */
14export function findTool(tools: readonly ToolInfo[], suffix: string): { server: string; tool: string } | undefined {
15 const hit = tools.find(t => t.mcp && bare(t.name) === suffix)
16 return hit && splitName(hit.name)
17}
18
19/** The tool input's `namespace` field when it is a string, else ''. */
20export const namespaceOf = (input: unknown): string => {
21 const ns = (input as { namespace?: unknown } | null)?.namespace
22 return typeof ns === 'string' ? ns : ''
23}
24hooks/screen.ts 261 lines1/**
2 * Pure text screening for the AgentDB mod (ADR-445). Two jobs: find secrets (so none is stored) and find prompt-injection phrasing (so
3 * retrieved memory cannot instruct the model). Findings are NAMES only: the matched text is never returned, logged or counted by value.
4 */
5
6// BEGIN SHARED SCREEN (generated from plugins/ruflo-agentdb/hooks/screen.ts by scripts/sync-mod-screen.mjs; do not edit in a copy)
7export type Rules = readonly (readonly [string, RegExp])[]
8
9/** The secret shapes every mod screens for. A plugin adds its own after these, outside the markers. */
10export const COMMON_SECRETS: Rules = [
11 ['private key', /-----BEGIN [A-Z ]*PRIVATE KEY-----/],
12 ['aws access key', /\b(?:AKIA|ASIA)[0-9A-Z]{16}\b/],
13 ['github token', /\b(?:gh[pousr]_[A-Za-z0-9]{30,}|github_pat_[A-Za-z0-9_]{40,})\b/],
14 ['slack token', /\bxox[abprs]-[A-Za-z0-9-]{10,}/],
15 ['slack webhook', /\bhooks\.slack\.com\/services\/T[A-Z0-9]{6,}\/B[A-Z0-9]{6,}\/[A-Za-z0-9]{16,}/],
16 ['google api key', /\bAIza[0-9A-Za-z_-]{35}\b/],
17 ['anthropic or openai key', /\bsk-(?:(?:ant|proj|svcacct|admin)-[A-Za-z0-9_-]{20,}|(?=[A-Za-z]{0,40}\d)[A-Za-z0-9]{32,})/],
18 ['stripe key', /\b[rs]k_live_[A-Za-z0-9]{16,}/],
19 ['npm token', /\bnpm_[A-Za-z0-9]{36}\b/],
20 ['huggingface token', /\bhf_[A-Za-z0-9]{30,}\b/],
21 ['sendgrid key', /\bSG\.[A-Za-z0-9_-]{16,}\.[A-Za-z0-9_-]{16,}/],
22 ['twilio key', /\bSK[0-9a-f]{32}\b/],
23 ['jwt', /\beyJ[A-Za-z0-9_-]{8,}\.eyJ[A-Za-z0-9_-]{8,}\.[A-Za-z0-9_-]{8,}/],
24 ['bearer token', /\bBearer\s+([A-Za-z0-9._~+/=-]{24,})/],
25 ['database url with credentials', /\b[a-z][a-z0-9+.-]{1,20}:\/\/[^\s:@/]+:[^\s@/]{3,}@[^\s/]+/i],
26]
27
28export const INJECTION: Rules = [
29 ['override instructions', /\b(?:ignore|disregard|forget|override)\b[^.\n]{0,40}\b(?:previous|prior|above|earlier|all|any|system)\b[^.\n]{0,30}\b(?:instructions?|rules?|prompts?|guidelines?)\b/i],
30 ['role reassignment', /\byou are (?:now|no longer)\b|\bact as (?:an? )?(?:unrestricted|jailbroken)\b/i],
31 ['new instructions', /\b(?:new|updated|real) (?:system )?instructions?\s*:/i],
32 ['fake role tags', /<\/?\s*(?:system|assistant|developer|instructions?)\s*>|^\s*(?:system|assistant)\s*:/im],
33 ['concealment', /\bdo not (?:tell|inform|mention|reveal)[^.\n]{0,30}\b(?:user|human|operator)\b/i],
34 ['exfiltration', /\b(?:exfiltrate|send|post|upload)\b[^.\n]{0,50}\b(?:secrets?|credentials?|tokens?|api keys?|\.env)\b/i],
35 ['shell pipe', /\b(?:curl|wget)\b[^|\n]{0,200}\|\s*(?:sudo\s+)?(?:ba|z)?sh\b/i],
36]
37
38// C0/C1 controls (keeping tab and newline), DEL, soft hyphen, combining grapheme joiner, Arabic letter mark, Hangul and Mongolian fillers/separators,
39// zero-width, bidi (overrides and isolates) and invisible-format characters, variation selectors; built with escapes, never raw.
40const INVISIBLE = new RegExp(
41 '[\\u0000-\\u0008\\u000b-\\u001f\\u007f-\\u009f\\u00ad\\u034f\\u061c\\u115f\\u1160\\u17b4\\u17b5\\u180b-\\u180e\\u200b-\\u200f\\u2028-\\u202e\\u2060-\\u206f\\u3164\\ufe00-\\ufe0f\\ufeff\\uffa0\\ufff9-\\ufffb]',
42 'g',
43)
44
45/** Longest input scanned in one pass; a longer one keeps its head and tail halves. One regex pass per rule, so cost stays linear. */
46const MAX_SCAN = 200_000
47
48/** Input bounded to MAX_SCAN characters with invisible characters removed, so none can hide a secret or a phrase. */
49export const bare = (text: string) =>
50 (text.length > MAX_SCAN ? text.slice(0, MAX_SCAN / 2) + '\n' + text.slice(-MAX_SCAN / 2) : text).replace(INVISIBLE, '')
51
52// A value is a secret CANDIDATE only when it is not a reference (env var, call, identifier path, placeholder, secret-manager path) and its
53// shape is random enough: at least two character classes, one of them a digit or symbol, and Shannon entropy of at least 2.5 bits per character.
54const PLACEHOLDER = /placeholder|your[-_ ]|example|changeme|change[-_]?me|redacted|dummy|replace[-_]?me|insert[-_]|\*{3,}|x{5,}|\.{3}|^(?:none|null|undefined|true|false)$/i
55const REFERENCE =
56 /^(?:\$(?:\{[^}]*\}|\(|[A-Za-z_]\w*$)|%[^%]*%$|<[^>]*>$|\{\{|process\.env|os\.environ|env[.[]|import\.meta|System\.getenv|secrets?\.|vault:|op:\/\/|ref\+|arn:|projects\/[^/]+\/secrets\/|gcp:|kms:|aws:|file:)/i
57const CALL = /^[A-Za-z_$][\w$]*(?:\.[A-Za-z_$][\w$]*)*\(|^[A-Za-z_$][\w$]*\[/
58const IDENT_PATH = /^[A-Za-z_$][\w$]*(?:\.[A-Za-z_$][\w$]*)+$/
59const UUID = /^[0-9a-f]{8}-(?:[0-9a-f]{4}-){3}[0-9a-f]{12}$/i
60const NAME_LIKE = /^[a-z][a-z0-9]*(?:[-_./][a-z0-9]+){2,}$/
61
62function entropy(v: string): number {
63 const counts = new Map<string, number>()
64 for (const ch of v) counts.set(ch, (counts.get(ch) ?? 0) + 1)
65 let h = 0
66 for (const n of counts.values()) h -= (n / v.length) * Math.log2(n / v.length)
67 return h
68}
69
70/** True when `v` is a name, call, path or placeholder rather than a literal credential. */
71function isReference(v: string): boolean {
72 if (PLACEHOLDER.test(v) || REFERENCE.test(v) || CALL.test(v) || IDENT_PATH.test(v) || URL_NO_CREDS.test(v)) return true
73 return NAME_LIKE.test(v) && v.replace(/\D/g, '').length / v.length < 0.15
74}
75
76export function plausibleSecret(v: string): boolean {
77 if (v.length < 8 || v.length > 256 || /\s/.test(v) || UUID.test(v) || isReference(v)) return false
78 const symbol = /[^A-Za-z0-9]/.test(v)
79 const digit = /\d/.test(v)
80 const classes = [/[a-z]/.test(v), /[A-Z]/.test(v), digit, symbol].filter(Boolean).length
81 return classes >= 2 && (digit || symbol) && entropy(v) >= 2.5
82}
83
84/** Under a secret-named key a literal this long is a secret even with one character class or a UUID shape, unless it is a clear reference. */
85const KEYED_MIN = 20
86const PLACEHOLDER_WORD = /(?:^|[^a-z])(?:your|placeholder|changeme|change[-_]?me|example|redacted|dummy|replace[-_]?me|insert)(?:[^a-z]|$)|\*{3,}|x{5,}|\.{3}|^(?:none|null|undefined|true|false)$/i
87const URL_NO_CREDS = /^[a-z][a-z0-9+.-]{1,20}:\/\/[^\s@]*$/i
88
89/** Three or more lowercase hyphen-separated words (no hex or digit-only run of 8+, few digits), such as my-k8s-secret-name-for-database. */
90function hyphenName(v: string): boolean {
91 const parts = v.split('-')
92 return parts.length >= 3 && parts.every(p => /^[a-z0-9]{2,}$/.test(p) && !/^[0-9a-f]{8,}$/.test(p)) && v.replace(/\D/g, '').length / v.length < 0.15
93}
94
95function keyedSecret(v: string): boolean {
96 if (v.length < KEYED_MIN || v.length > 256 || /\s/.test(v)) return false
97 return !(PLACEHOLDER_WORD.test(v) || REFERENCE.test(v) || CALL.test(v) || IDENT_PATH.test(v) || URL_NO_CREDS.test(v) || (!UUID.test(v) && hyphenName(v)))
98}
99
100const KEY_NAME = /(?:api[_-]?key|secret|token|passw(?:or)?d|passwd|pwd|credential|private[_-]?key|auth(?!or))s?[A-Za-z0-9_-]{0,40}["']?\s*[:=]\s*/gi
101const QUOTED = /(["'\x60])((?:(?!\1)[^\n]){1,256})\1/y
102const BARE_VALUE = /[^\s"'\x60,;]{1,256}/y
103const QUERY_VALUE = /[^\s"'\x60,;&]{1,256}/y
104
105/** A secret-named key assigned a literal value: env style, JSON, YAML, code. Values that are calls, references or placeholders do not count. */
106function assignmentSecret(text: string): boolean {
107 let valueEnd = 0
108 for (const m of text.matchAll(KEY_NAME)) {
109 if (m.index < valueEnd) continue // a key-looking word inside the previous value, such as secretsmanager in an ARN
110 const at = m.index + m[0].length
111 let back = m.index
112 while (back > 0 && m.index - back < 64 && /[A-Za-z0-9_.-]/.test(text.charAt(back - 1))) back--
113 const re = /["'\x60]/.test(text.charAt(at)) ? QUOTED : /[?&]/.test(text.charAt(back - 1)) ? QUERY_VALUE : BARE_VALUE
114 re.lastIndex = at
115 const hit = re.exec(text)
116 const v = hit && (hit[2] ?? hit[0])
117 valueEnd = hit ? at + hit[0].length : at
118 if (v && (plausibleSecret(v) || keyedSecret(v))) return true
119 }
120 return false
121}
122
123/** A password in a URL's userinfo that is not a placeholder such as user:password or ${DB_PASSWORD}. */
124function urlCredential(url: string): boolean {
125 const pass = /^[^:]+:\/\/[^\s:@/]+:([^\s@/]+)@/.exec(url)?.[1]
126 if (!pass || /\$\{|\{\{|%\(|%s/.test(pass)) return false
127 return !/^(?:password|passwd|pass|pwd|secret|changeme|dbpassword|db_password|\$\w*|<.*>|\{.*\}|\*+|x+)$/i.test(pass) && !PLACEHOLDER.test(pass)
128}
129
130const CHECKS: Readonly<Record<string, (m: RegExpMatchArray) => boolean>> = {
131 'bearer token': m => !isReference(m[1] ?? ''),
132 'database url with credentials': m => urlCredential(m[0]),
133 'database url with password': m => urlCredential(m[0]),
134}
135const globals = new WeakMap<RegExp, RegExp>()
136
137function matches(name: string, re: RegExp, text: string): boolean {
138 if (name === 'key assignment') return assignmentSecret(text)
139 const check = CHECKS[name]
140 if (!check) return re.test(text)
141 let g = globals.get(re)
142 if (!g) globals.set(re, (g = new RegExp(re.source, re.flags.includes('g') ? re.flags : re.flags + 'g')))
143 for (const m of text.matchAll(g)) if (check(m)) return true
144 return false
145}
146
147/**
148 * The text textsOf appends when it had to drop input (a node, character or per-string budget ran out). It is never matched against a rule:
149 * `names` reports it as a finding of its own, so every guard that asks "is there a secret in these texts" refuses what it could not read in full.
150 */
151export const TRUNCATED = 'ruflo-screen: input exceeded the screening budget'
152export const TRUNCATED_NAME = 'input too large to screen'
153
154/** Names of the rules that match `text` (already bare'd). A rule named 'key assignment' is judged by assignmentSecret, whatever its regex. */
155export const names = (rules: Rules, text: string) => text === TRUNCATED ? [TRUNCATED_NAME] : rules.filter(([name, re]) => matches(name, re, text)).map(([name]) => name)
156
157export type Findings = { readonly secrets: readonly string[]; readonly injection: readonly string[] }
158
159/** Names of every secret shape in `secrets` and every injection phrase found in `text`. Cost is linear in the capped input. */
160export function screenWith(secrets: Rules, text: string): Findings {
161 const bounded = bare(text)
162 return { secrets: names(secrets, bounded), injection: names(INJECTION, bounded) }
163}
164
165export const hasSecretIn = (secrets: Rules, text: string) => names(secrets, bare(text)).length > 0
166
167/** Makes stored text safe to show: no control or bidi characters, whitespace collapsed, at most `max` characters. */
168export function tidy(text: string, max: number): string {
169 const flat = text.replace(INVISIBLE, '').replace(/\s+/g, ' ').trim()
170 return flat.length > max ? `${flat.slice(0, Math.max(0, max - 1))}…` : flat
171}
172/** Bounds for textsOf: nodes visited, characters returned, the longest string read in full, the size of one returned chunk, chunk overlap. */
173export type TextLimits = { readonly nodes?: number; readonly chars?: number; readonly perString?: number }
174const NODES = 20_000
175const CHARS = 2_000_000
176const PER_STRING = 1_500_000
177const OVERLAP = 2_048
178const BARE_KEY_MIN = 8
179
180/** A string as texts the screen can read whole: one text up to MAX_SCAN, else overlapping MAX_SCAN windows so a secret anywhere is inside one. */
181function windows(text: string, out: string[]): void {
182 if (text.length <= MAX_SCAN) {
183 out.push(text)
184 return
185 }
186 for (let at = 0; ; at += MAX_SCAN - OVERLAP) {
187 out.push(text.slice(at, at + MAX_SCAN))
188 if (at + MAX_SCAN >= text.length) return
189 }
190}
191
192/**
193 * Every string in a tool input, for the screen to read: iterative (no recursion, so nesting 5000 deep cannot overflow the stack) and
194 * breadth-first (siblings before depth, so a long list cannot hide a nested value). A string under an object key comes back as `key=value`,
195 * so a secret-named key is judged with its value; a key whose value is not a string is returned bare. Strings longer than the screen window
196 * come back as overlapping windows; one over `perString` keeps its head and tail. Work is bounded by `nodes` slots and `chars` characters.
197 * Anything dropped (slots or characters ran out, or a string lost its middle) is reported by a final TRUNCATED text, which `names` and
198 * `hasSecretIn` count as a finding, so the screen fails closed instead of passing what it did not read.
199 */
200export function textsOf(input: unknown, limits: TextLimits = {}): string[] {
201 const out: string[] = []
202 let slots = limits.nodes ?? NODES
203 let chars = limits.chars ?? CHARS
204 const perString = limits.perString ?? PER_STRING
205 let truncated = false
206 const take = (text: string): void => {
207 if (chars <= 0) {
208 truncated = true
209 return
210 }
211 if (text.length <= Math.min(perString, chars)) {
212 chars -= text.length
213 windows(text, out)
214 return
215 }
216 truncated = true
217 const half = Math.floor(Math.min(perString, chars) / 2)
218 chars -= 2 * half
219 windows(text.slice(0, half), out)
220 windows(text.slice(-half), out)
221 }
222 const queue: unknown[] = [input]
223 let head = 0
224 for (; head < queue.length && chars > 0; head++) {
225 const node = queue[head]
226 if (typeof node === 'string') take(node)
227 else if (Array.isArray(node)) {
228 let i = 0
229 for (; i < node.length && slots > 0; i++, slots--) if (i in node) queue.push(node[i])
230 if (i < node.length) truncated = true
231 } else if (typeof node === 'object' && node !== null) {
232 for (const k in node) {
233 if (slots-- <= 0) {
234 truncated = true
235 break
236 }
237 if (!Object.prototype.hasOwnProperty.call(node, k)) continue
238 const v = (node as Record<string, unknown>)[k]
239 if (typeof v === 'string') queue.push(k + '=' + v)
240 else {
241 if (k.length >= BARE_KEY_MIN) take(k)
242 queue.push(v)
243 }
244 }
245 }
246 }
247 if (queue.length > head) truncated = true
248 if (truncated) out.push(TRUNCATED)
249 return out
250}
251// END SHARED SCREEN
252
253const SECRETS: Rules = [
254 ...COMMON_SECRETS,
255 ['key assignment', /\b(?:api[_-]?key|secret|token|passw(?:or)?d|credential)s?["']?\s*[:=]\s*["']?[A-Za-z0-9/+=_.-]{16,}/i],
256]
257
258export const scan = (text: string): Findings => screenWith(SECRETS, text)
259
260export const hasSecret = (text: string) => hasSecretIn(SECRETS, text)
261