registry / AI-agent capability abuse / instar / 1.3.1322

instar@1.3.1322

Coherence infrastructure for self-evolving AI agents — on the Claude Code or Codex subscription you already have.

AI Security Review

scanned 18d ago · by lpm-firewall-ai

The setup flow creates a concrete agent configuration capability by registering a remotely fetched MCP server and accepting Claude project trust. Its trigger is the interactive setup command, while npm lifecycle scripts do not reach this behavior, so this warrants a warning rather than a block.

Static reason

High-risk behavior combination matched malicious policy.; previous stored version diff introduced dangerous source

Trigger

Running the Instar setup flow.

Impact

The configured agent can launch the Playwright MCP server without the usual project trust prompt.

Mechanism

Setup adds the Playwright MCP command and arguments to agent configuration and marks the Claude project trust dialog accepted.

AI rationale

Evidence

dist/commands/setup.js

Decision evidence

public snapshot

AI called this Suspicious at 91.0% confidence as Dangerous Capability with medium false-positive risk.

Evidence for warning

The interactive setup flow registers Playwright MCP and pre-accepts Claude project trust, allowing the server to load without a trust dialog.
The MCP registration uses npx to fetch @playwright/mcp@latest when launched.

Evidence against

The npm postinstall runs a native dependency repair script; it does not invoke the agent configuration setup.
The agent configuration write is reached from the setup flow, not from an npm install hook.

Behavioral surface

Source

ChildProcessCryptoDynamicRequireEnvironmentVarsFilesystemNetworkShellWebSocket

Supply chain

HighEntropyStringsPossibleObfuscationUrlStrings

ManifestNo manifest risk signals triggered.

scanned 2,020 file(s), 29.1 MB of source, external domains: 127.0.0.1, abc123.trycloudflare.com, accounts.google.com, ai.google.dev, api.anthropic.com, api.cloudflare.com, api.github.com, api.groq.com, api.mail.tm, api.openai.com, api.slack.com, api.telegra.ph, api.telegram.org, api.typesafe.ai, auth.anthropic.com, auth.openai.com, blob.vercel-storage.com, bot-me.ai, brew.sh, chatgpt.com, claude.ai, claude.com, cli.github.com, console.anthropic.com, console.cloud.google.com, dawn.bot-me.ai, docs.anthropic.com, example.invalid, feedback.dawn-tunnel.dev, files.slack.com, git-scm.com, github.com, graph.facebook.com, groq.com, instar-telemetry.sagemind-ai.workers.dev, instar.sh, myaccount.google.com, myapp.vercel.app, nodejs.org, openrouter.ai, platform.claude.com, platform.openai.com, raw.githubusercontent.com, sagemindai.io, sentry.io, slack.com, staging.myapp.com, support.google.com, telegra.ph, threadline-relay.fly.dev

Source & flagged code

21 flagged

package.jsonView file

Lines 44-84json

44 "lint:undefined-erasing-default-merge": "node scripts/lint-no-undefined-erasing-default-merge.js",

45 "lint:destructive": "node scripts/lint-no-direct-destructive.js",

46 "lint:dev-agent-dark-gate": "node scripts/lint-dev-agent-dark-gate.js",

47 "lint:guard-manifest": "node scripts/lint-guard-manifest.js",

48 "lint:dev-agent-dark-gate:staged": "node scripts/lint-dev-agent-dark-gate.js --staged",

49 "lint:destructive:staged": "node scripts/lint-no-direct-destructive.js --staged",

50 "lint:llm-http": "node scripts/lint-no-direct-llm-http.js",

51 "lint:scrape-realness": "node scripts/lint-scrape-fixture-realness.js",

52 "lint:scrape-realness:staged": "node scripts/lint-scrape-fixture-realness.js --staged",

53 "lint:llm-attribution": "node scripts/lint-llm-attribution.js",

54 "lint:llm-attribution:staged": "node scripts/lint-llm-attribution.js --staged",

55 "lint:llm-http:staged": "node scripts/lint-no-direct-llm-http.js --staged",

56 "lint:credential-write": "node scripts/lint-no-unfunneled-credential-write.js",

57 "lint:credential-write:staged": "node scripts/lint-no-unfunneled-credential-write.js --staged",

58 "clean": "rm -rf dist",

59 "check:upgrade-guide": "node scripts/check-upgrade-guide.js",

60 "check:pre-push-gate": "node scripts/pre-push-gate.js",

61 "check:release": "node scripts/analyze-release.js",

62 "check:contract-evidence": "node scripts/check-contract-evidence.js",

63 "prepublishOnly": "npm run build && node scripts/verify-codex-stage-b-release-evidence.mjs && npm run check:upgrade-guide && npm run check:contract-evidence",

64 "postinstall": "node scripts/fix-better-sqlite3.cjs",

High

Install Time Lifecycle Scripts

Package defines install-time lifecycle scripts.

package.jsonView on unpkg · L64

65 "prepare": "husky",

66 "lint:standards-registry-path": "node scripts/lint-no-direct-standards-registry-path.mjs",

67 "lint:framework-list": "node scripts/lint-framework-list-completeness.mjs"

68 },

69 "keywords": [

70 "instar",

71 "ai-agent",

72 "autonomous-agent",

73 "claude",

74 "claude-code",

75 "agent-framework",

76 "telegram-bot",

77 "job-scheduler",

78 "agent-memory",

79 "agent-identity",

80 "llm",

81 "mcp",

82 "cli",

83 "molt",

84 "lobster"

skills/credential-leak-detector/SKILL.mdView file

Lines 7-47markdown

7 version: "1.0"

8 homepage: https://instar.sh

9---

11# credential-leak-detector — Catch Leaked Credentials Before They Spread

13Every time your agent runs a Bash command, the output flows back into the conversation — and into your API provider's logs, any monitoring tools, and the model's context window. If that output contains an API key, a private key, or a database password, the credential is now exposed in places you never intended.

15This hook scans every Bash tool response for 14 credential patterns before the output reaches the agent. Critical matches (API keys, AWS credentials, private keys) get blocked entirely. High-severity matches get redacted with a warning. Suspicious patterns get flagged as advisories. No external dependencies — just Python stdlib.

dist/threadline/PipeSessionSpawner.jsView file

Lines 1-35javascript

1/**

2 * PipeSessionSpawner — Spawns lightweight `claude -p` sessions for simple

3 * threadline queries. Sessions auto-exit when done, reducing zombie risk

4 * and session slot consumption.

5 *

6 * Part of RFC: Persistent Listener Daemon Architecture (Phase 2).

7 *

8 * Security model:

9 * - Messages wrapped in <untrusted-message> XML tags (injection-hardened)

10 * - Thread history LLM-summarized before injection (strips jailbreak assembly)

scripts/check-rule3-coverage.cjsView file

Lines 16-56javascript

16 * - tmux capture-pane / send-keys patterns

17 * - JSON.parse on subprocess stdout

18 * - new class

dist/commands/listener.jsView file

Lines 1-34javascript

1/**

2 * `instar listener` — Manage the standalone listener daemon.

3 *

4 * Commands:

5 * instar listener start — Start the listener daemon

6 * instar listener stop — Gracefully stop the daemon

7 * instar listener status — Show daemon state + connection info

8 * instar listener logs — Tail daemon log file

9 * instar listener restart — Graceful restart

10 * instar listener doctor — Pre-flight check

skills/spec-converge/scripts/publish-spec-review.mjsView file

Lines 1-40javascript

1#!/usr/bin/env node

2/**

3 * publish-spec-review.mjs — the sanctioned way to deliver a spec for review.

4 *

5 * Structure > Willpower: sending a spec for review WITHOUT a rendered,

6 * verified ELI16 tunnel link should be impossible, not something the agent

7 * remembers to attach. This script makes the correct delivery atomic:

8 *

9 * 1. Resolve the spec's ELI16 companion using the SAME convention the

10 * commit-time gate enforces (scripts/eli16-overview-check.mjs).

dist/messaging/TelegramAdapter.jsView file

•Manifest-reachable source captures an API credential, sends it to a fixed unofficial gateway, and persists that redirection in agent or shell configuration. dist/messaging/TelegramAdapter.js: fs.writeFileSync(tmpPath, JSON.stringify(config, null, 2)); fs.writeFileSync(tmpPath, JSON.stringify({ messageId, savedAt: new Date().toISOString() })); fs.writeFileSync(tmpPath, JSON.stringify(config, null, 2)); * Public read-only authorization check for a sender id (Know Your Principal #898, fs.writeFileSync(destPath, buffer); baseUrl: 'https://api.groq.com/openai/v1', baseUrl: 'https://api.openai.com/v1', return { apiKey, baseUrl: p.baseUrl, model: p.model };

High

Credential Redirect Persistence

Manifest-reachable source sends a prompted API credential to a fixed unofficial gateway and persists the redirection.

dist/messaging/TelegramAdapter.jsView on unpkg

dist/core/GitSync.jsView file

Lines 1-32javascript

1/**

2 * Git-based state synchronization for multi-machine coordination.

3 *

4 * Handles:

5 * - Configuring git commit signing with machine Ed25519 keys

6 * - Commit signing verification on pull

7 * - Debounced auto-commit + push

8 * - Relationship merge with field-level resolution

9 * - Conflict resolution strategies

10 *

11 * Part of Phase 3 (state sync via git).

dist/commands/setup.jsView file

•matchType = previous_version_dangerous_delta matchedPackage = instar@1.3.1315 matchedIdentity = npm:aW5zdGFy:1.3.1315 similarity = 0.933 summary = stored previous version shares package body but lacks this dangerous source file

Critical

Previous Version Dangerous Delta

This package version adds a dangerous source file absent from the previous stored version; route for source-aware review.

dist/commands/setup.jsView on unpkg

17*/ L18: import { execFileSync, execSync, spawn } from 'node:child_process'; L19: import { telegramFetch } from '../messaging/telegram-egress.js'; ... L25: import pc from 'picocolors'; L26: const __dirname = path.dirname(fileURLToPath(import.meta.url)); L27: /** ... L105: console.log(); L106: const rl = readline.createInterface({ input: process.stdin, output: process.stdout }); L107: try { ... L215: console.log(pc.dim(' GitHub CLI (gh) not found. To discover cloud-backed agents:')); L216: const platform = process.platform; L217: if (platform === 'darwin') {

High

Sandbox Evasion Gated Capability

Source gates dangerous network, credential, or execution behavior behind CI, host, platform, time, or geo fingerprint checks.

dist/commands/setup.jsView on unpkg · L17

dist/core/PostUpdateMigrator.jsView file

•path = dist/core/PostUpdateMigrator.js kind = trigger_reachable_excluded_payload sizeBytes = 1304493 triggerReachableSource = dist/core/PostUpdateMigrator.js excludedPayload = .claude/hooks/free-text-guard.sh

Critical

Trigger Reachable Excluded Directory Payload

A package entrypoint or install-reachable source explicitly loads and activates an executable-looking payload from a test, fixture, or hidden path.

dist/core/PostUpdateMigrator.jsView on unpkg

•Manifest-trigger-reachable source links an external AI-agent control path to a behavior-bearing write operation. w); if (current.includes(anyrow) && start >= 0 && end > start) { current = current.replace(anyrow, subscription_signin_skill_content.slice(start, end) + anyrow); changes.push('three failure-table rows'); } } if (changes.length === 0) { if (!current.includes(marker) || !current.includes(rowsmarker)) { result.skipped.push('skills/subscription-signin/skill.md: customized — left untouched (no cross-machine section)'); } return; } fs.writefilesync(skillfile, current); result.upgraded.pus

High

Trigger Reachable External Ai Agent Control Surface Mutation

Manifest-trigger-reachable source writes behavior-bearing configuration into a user or project AI-agent control surface.

dist/core/PostUpdateMigrator.jsView on unpkg

scripts/run-contract-tests.jsView file

39try { L40: execSync('npx vitest run --config vitest.contract.config.ts', { L41: cwd: ROOT,

High

Runtime Package Install

Package source invokes a package manager install command at runtime.

scripts/run-contract-tests.jsView on unpkg · L39

.claude/hooks/free-text-guard.shView file

•path = .claude/hooks/free-text-guard.sh kind = payload_in_excluded_dir sizeBytes = 3888 magicHex = [redacted]

High

Payload In Excluded Dir

Package hides binary, compressed, or executable-looking payloads in test/fixture/hidden paths.

.claude/hooks/free-text-guard.shView on unpkg

•path = .claude/hooks/free-text-guard.sh kind = build_helper sizeBytes = 3888 magicHex = [redacted]

Medium

Ships Build Helper

Package ships non-JavaScript build or shell helper files.

.claude/hooks/free-text-guard.shView on unpkg

dist/commands/devPreflight.js#virtual:normalized:round1View file

•stage = ast_semantic_analysis; reason = ast_parse_error; limitedFiles = 81

High

Semantic Analysis Limited

A bounded semantic-analysis stage reached its safety limit; remaining detectors completed, but this package requires AI review.

dist/commands/devPreflight.js#virtual:normalized:round1View on unpkg

Source & flagged code

21 flagged · loading source

package.jsonView file

•scripts.postinstall = node scripts/fix-better-sqlite3.cjs

High

Install Time Lifecycle Scripts

Package defines install-time lifecycle scripts.

package.jsonView on unpkg

skills/credential-leak-detector/SKILL.mdView file

27patternName = aws_access_key severity = critical line = 27

Critical

Critical Secret

Package contains a critical-looking secret pattern.

skills/credential-leak-detector/SKILL.mdView on unpkg · L27

27patternName = aws_access_key severity = critical line = 27

Critical

Secret Pattern

AWS access key ID in skills/credential-leak-detector/SKILL.md

skills/credential-leak-detector/SKILL.mdView on unpkg · L27

Findings

7 Critical11 High6 Medium5 Low

CriticalCritical Secretskills/credential-leak-detector/SKILL.md

CriticalPersistence Backdoordist/core/GitSync.js

CriticalTrigger Reachable Dangerous Capabilitydist/core/GitSync.js

CriticalTrigger Reachable Excluded Directory Payloaddist/core/PostUpdateMigrator.js

CriticalPrevious Version Dangerous Deltadist/commands/setup.js

CriticalSecret Patternskills/credential-leak-detector/SKILL.md

HighInstall Time Lifecycle Scriptspackage.json

HighChild Processdist/threadline/PipeSessionSpawner.js

HighShelldist/threadline/PipeSessionSpawner.js

High

Affected versions and remediation

This report applies to instar@1.3.1322.

See all versions of instar for other recorded verdicts.

Review the evidence and your use of instar@1.3.1322 before allowing it. Restrict the permissions described in this report.
Choose an independently verified alternative or release. This report does not establish that other versions are safe.

Evidence last updated: 2026-10-07 02:55:10.655Z (UTC).

Think this verdict is wrong? Dispute it on GitHub or read how reports are produced.

AI Security Review

Decision evidence

Source & flagged code

Source & flagged code

Findings

Affected versions and remediation

Runtime surface

Behavioral surface

Source & flagged code