Back to skill

Security audit

vibe-ledger

Security checks for vulnerabilities and agentic risk

Overview

The skill is rated suspicious with medium confidence: scanners report that raw secrets may be saved in run files and that privacy disclosures, extra settings changes, and some test-fixture dependencies raise concerns.

Review the secret-handling and privacy concerns before installing, especially if you will analyze sensitive code or conversations. The scan was partial and some findings concern test fixtures or documentation, so the evidence is incomplete; the suspicious verdict does not establish malicious intent.

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (94)

Known Vulnerable Dependency: next==15.0.0 — 38 advisory(ies): GHSA-2464-fhr4-w9xw (Next.js: Unauthenticated Remote Code Execution in Image Optimization API when SV); GHSA-2xp9-vwfh-vxw4 (Next.js: Unauthenticated Remote Code Execution in Image Optimization API when AV); CVE-2026-44573 (Next.js has a Middleware / Proxy bypass in Pages Router applications using i18n) +35 more

Critical
Category
Supply Chain
Confidence
90% confidence
Finding

The fixture pins an obsolete Next.js release with known security vulnerabilities, creating exposure if someone installs and serves it. Its role as static audit input substantially reduces the skill's immediate risk, and this manifest alone does not establish applicability of every listed advisory or the claimed critical RCE paths.

Content

No source excerpt is available for this finding.

Known Vulnerable Dependency: next==15.0.0 — 38 advisory(ies): GHSA-2464-fhr4-w9xw (Next.js: Unauthenticated Remote Code Execution in Image Optimization API when SV); GHSA-2xp9-vwfh-vxw4 (Next.js: Unauthenticated Remote Code Execution in Image Optimization API when AV); CVE-2026-44573 (Next.js has a Middleware / Proxy bypass in Pages Router applications using i18n) +35 more

Critical
Category
Supply Chain
Confidence
94% confidence
Finding

The fixture pins an outdated Next.js release with known security vulnerabilities, although this manifest does not establish applicability of every advisory listed by the scanner. Because it is a test fixture rather than a demonstrated skill runtime dependency, exploitation would require separately installing and running the sample application with affected features exposed.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The supplied code implements a chat backend, not the described project-explanation and requirements-auditing tool. It contains no file analysis, conversation auditing, requirements verification, or HTML report generation. The external OpenAI request directly conflicts with the offline-only claim. The file path identifies this as a test fixture, so it does not by itself establish that the skill executes this server; nevertheless, the behavior of the supplied chunk does not match the declared purpose.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The supplied chunk implements a note-storage API, not the declared project-explanation and verification functionality. It contains no project analysis, conversation auditing, flowchart generation, or HTML reporting. Its path identifies it as a test fixture, so it could legitimately support the larger skill's tests; this mismatch assessment applies to the supplied chunk's behavior, not proof that the overall skill performs undeclared actions. The code does not show outbound data transmission.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

Most visible behavior supports the declared purpose. However, do_retention with --set writes to Claude Code's settings.json, or a supplied --settings path, and backs up an existing configuration. The start workflow also recommends setting retention to 365 days after user agreement. Modifying application-wide transcript-retention policy is a substantive capability not disclosed in the description, rather than merely an implementation detail of generating a report. This change requires a separate command; it is not performed automatically by start. No outbound transmission is visible in this chunk, though the imported modules are not supplied for inspection.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The call to openai.chat.completions.create sends JSON.stringify(notes) to an external service. This directly conflicts with the declared local-only behavior and serves a materially different purpose. The path identifies this as a test-fixture application file, so the snippet alone does not establish that the skill itself invokes this endpoint; the supplied code's behavior nevertheless differs from the declared purpose.

Content

No source excerpt is available for this finding.

Agent Config Directory Access

High
Category
Agent Snooping
Confidence
90% confidence
Finding

Skill reads from agent configuration directories (.claude/, .codex/, .gemini/). These directories may contain API keys, personal settings, and other credentials that the skill has no legitimate need to access.

Content

Scanner excerpt · docs/superpowers/plans/2026-10-10-vibe-ledger.md (reported line 20)May include surrounding context.

md
- Intermediate files live in `~/.vibe-ledger/projects/<project-id>/runs/<timestamp>/` (overridable with `VIBE_LEDGER_HOME`). Project id = `slug(folder name)-<first 8 hex of SHA-1 of the resolved absolute path>`.
- Reports go to `~/Desktop`, else `~/Downloads`, else the run folder (`VIBE_LEDGER_OUT_DIR` or `--out-dir` override).
- Secrets are masked (prefix + `…` + last 4) in the digest, the archive and the report; the renderer refuses to write a report that still contains one.
- `~/.claude/settings.json` is changed only by `vl.py retention --set`, which SKILL.md tells the agent to run only after the user agrees.
- Constants (copied from the spec): reading-pack limit **120,000** characters; compromise window **20** digest records or the end of the next turn, whichever is earlier; tool-error text truncated to **2,000** characters; edit summaries to **1,500**; code-snippet tolerance **±3** lines; map shows at most **12** nodes; high-entropy secret = length ≥ **20** and Shannon entropy ≥ *

Agent Config Directory Access

High
Category
Agent Snooping
Confidence
90% confidence
Finding

Skill reads from agent configuration directories (.claude/, .codex/, .gemini/). These directories may contain API keys, personal settings, and other credentials that the skill has no legitimate need to access.

Content

Scanner excerpt · docs/superpowers/specs/2026-10-09-vibe-ledger-design.md (reported line 282)May include surrounding context.

md
- 全部在本机运行,不向外发送任何代码或对话。
- `redact.py` 对以下内容打码,只保留末 4 位:`sk-`、`sk-ant-`、`sk_live_`、`pk_live_`、`AKIA…`、`ghp_`、`xoxb-`、`xoxp-`、JWT(`eyJ…` 长串)、私钥块,以及赋值给名称含 `KEY`、`SECRET`、`TOKEN`、`PASSWORD` 的变量的高熵字符串(长度不少于 20,且香农熵不低于每字符 3.5 比特)。生成 digest 和渲染报告时**都要**过一遍。
- 中间文件和归档放在仓库外,避免被误提交。
- 修改用户的 `~/.claude/settings.json` 之前必须先征得同意。

## 13. 测试与验收

Credential Access

High
Category
Privilege Escalation
Confidence
60% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · docs/superpowers/plans/2026-10-10-vibe-ledger.md (reported line 2166)May include surrounding context.

md
def test_client_prefixed_secret_in_env_is_certain(tmp_path):
    write(tmp_path, ".env.local", f"NEXT_PUBLIC_SUPABASE_ANON_KEY={jwt('anon')}\nNEXT_PUBLIC_OPENAI_API_KEY={OPENAI}\n"
                                  "OPENAI_API_KEY_SERVER=fine-on-server\nVITE_ADMIN_SECRET=x\n")
    write(tmp_path, ".env.example", f"NEXT_PUBLIC_OPENAI_API_KEY={OPENAI}\n")
    found = secrets_exposed.run(tmp_path, {"stacks": ["nextjs"]})

Credential Access

High
Category
Privilege Escalation
Confidence
60% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · docs/superpowers/plans/2026-10-10-vibe-ledger.md (reported line 2171)May include surrounding context.

md
write(tmp_path, ".env.example", f"NEXT_PUBLIC_OPENAI_API_KEY={OPENAI}\n")
    found = secrets_exposed.run(tmp_path, {"stacks": ["nextjs"]})
    got = sorted((f.path, f.line, f.level) for f in found)
    assert got == [(".env.local", 2, "certain"), (".env.local", 4, "certain")]


def test_hardcoded_secret_in_code_but_not_anon_jwt(tmp_path):

Credential Access

High
Category
Privilege Escalation
Confidence
60% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · docs/superpowers/plans/2026-10-10-vibe-ledger.md (reported line 2171)May include surrounding context.

md
write(tmp_path, ".env.example", f"NEXT_PUBLIC_OPENAI_API_KEY={OPENAI}\n")
    found = secrets_exposed.run(tmp_path, {"stacks": ["nextjs"]})
    got = sorted((f.path, f.line, f.level) for f in found)
    assert got == [(".env.local", 2, "certain"), (".env.local", 4, "certain")]


def test_hardcoded_secret_in_code_but_not_anon_jwt(tmp_path):

Credential Access

High
Category
Privilege Escalation
Confidence
60% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · docs/superpowers/plans/2026-10-10-vibe-ledger.md (reported line 3976)May include surrounding context.

md
"E2": {"kind": "code", "path": "supabase/migrations/001.sql", "line": 3,
               "snippet": "alter table notes disable row level security;", "ok": True, "line_corrected": False},
        "E3": {"kind": "probe", "finding_id": "secrets_exposed:.env.local:2", "ok": True, "probe": "secrets_exposed",
               "level": "certain", "path": ".env.local", "line": 2,
               "snippet": "NEXT_PUBLIC_OPENAI_API_KEY=" + OPENAI, "message": "bundled into the browser"},
        "E4": {"kind": "transcript", "session": "s1", "msg_id": "m4", "quote": "暂时先把 key 放前端",
               "ok": True, "ts": "2026-09-13T10:00:00Z", "role": "assistant", "msg_kind": "text"},

Credential Access

High
Category
Privilege Escalation
Confidence
60% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · docs/superpowers/plans/2026-10-10-vibe-ledger.md (reported line 3996)May include surrounding context.

md
"fix_prompt": "Re-enable row level security on notes and add a policy using (auth.uid() = user_id)."},
        {"id": "P2", "type": "deferred", "question": "open_debts", "plain": "把 OpenAI key 移到服务端",
         "source": ["E4"], "status": "pending", "verdict_plain": "key 还在前端,任何人都能拿走",
         "verdict_dev": "NEXT_PUBLIC_OPENAI_API_KEY in .env.local", "evidence": ["E3"],
         "fix_prompt": "Rename NEXT_PUBLIC_OPENAI_API_KEY to OPENAI_API_KEY and read it only in route.ts."},
        {"id": "P3", "type": "ai_claim", "question": "what_costs", "plain": "接口已经加了登录保护",
         "source": ["E1"], "status": "unknown", "verdict_plain": "没找到能证明这一点的代码",

Credential Access

High
Category
Privilege Escalation
Confidence
60% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · docs/superpowers/plans/2026-10-10-vibe-ledger.md (reported line 6637)May include surrounding context.

md
"deterministic": {
    "findings": [
      {"probe": "supabase_rls", "path": "supabase/migrations/001_notes.sql", "level": "certain"},
      {"probe": "secrets_exposed", "path": ".env.local", "level": "certain"},
      {"probe": "paid_api_unauthed", "path": "app/api/summarize/route.ts", "level": "hint"}
    ],
    "no_certain_findings": false,

Credential Access

High
Category
Privilege Escalation
Confidence
60% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · docs/superpowers/plans/2026-10-10-vibe-ledger.md (reported line 6648)May include surrounding context.

md
{"label": "RLS switched off breaks 'only I can see my notes'", "status": ["broken"],
     "paths_any": ["supabase/migrations"], "text_any": ["rls", "row level", "行级", "访问规则"]},
    {"label": "the OpenAI key in a NEXT_PUBLIC_ variable is still there", "status": ["pending", "broken"],
     "paths_any": [".env.local", "app/api/summarize"], "text_any": ["key", "密钥", "openai"]},
    {"label": "the claimed login protection on /api/summarize does not exist", "status": ["broken", "partial"],
     "paths_any": ["app/api/summarize"], "text_any": ["登录", "login", "auth"]}
  ],

Credential Access

High
Category
Privilege Escalation
Confidence
60% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · docs/superpowers/plans/2026-10-10-vibe-ledger.md (reported line 6773)May include surrounding context.

md
s.user("再加一个 AI 总结按钮,不要让陌生人随便调用花我的钱。")
    s.assistant_text("我加一个 /api/summarize 接口。暂时先把 OpenAI key 放在 NEXT_PUBLIC_OPENAI_API_KEY 里,之后再移到服务端变量。")
    s.write(p / "app/api/summarize/route.ts", read("app/api/summarize/route.ts"))
    s.write(p / ".env.local", read(".env.local"))
    s.assistant_text("AI 总结已完成,接口已经加了登录保护,只有登录用户才能调用。")
    s.save(logs / "claude" / "projects")
    return "claude-code"

Credential Access

High
Category
Privilege Escalation
Confidence
60% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · docs/superpowers/plans/2026-10-10-vibe-ledger.md (reported line 6773)May include surrounding context.

md
s.user("再加一个 AI 总结按钮,不要让陌生人随便调用花我的钱。")
    s.assistant_text("我加一个 /api/summarize 接口。暂时先把 OpenAI key 放在 NEXT_PUBLIC_OPENAI_API_KEY 里,之后再移到服务端变量。")
    s.write(p / "app/api/summarize/route.ts", read("app/api/summarize/route.ts"))
    s.write(p / ".env.local", read(".env.local"))
    s.assistant_text("AI 总结已完成,接口已经加了登录保护,只有登录用户才能调用。")
    s.save(logs / "claude" / "projects")
    return "claude-code"

Credential Access

High
Category
Privilege Escalation
Confidence
60% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · tests/e2e/scenarios.py (reported line 41)May include surrounding context.

python
s.user("再加一个 AI 总结按钮,不要让陌生人随便调用花我的钱。")
    s.assistant_text("我加一个 /api/summarize 接口。暂时先把 OpenAI key 放在 NEXT_PUBLIC_OPENAI_API_KEY 里,之后再移到服务端变量。")
    s.write(p / "app/api/summarize/route.ts", read("app/api/summarize/route.ts"))
    s.write(p / ".env.local", read(".env.local"))
    s.assistant_text("AI 总结已完成,接口已经加了登录保护,只有登录用户才能调用。")
    s.save(logs / "claude" / "projects")
    return "claude-code"

Credential Access

High
Category
Privilege Escalation
Confidence
60% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · tests/e2e/scenarios.py (reported line 41)May include surrounding context.

python
s.user("再加一个 AI 总结按钮,不要让陌生人随便调用花我的钱。")
    s.assistant_text("我加一个 /api/summarize 接口。暂时先把 OpenAI key 放在 NEXT_PUBLIC_OPENAI_API_KEY 里,之后再移到服务端变量。")
    s.write(p / "app/api/summarize/route.ts", read("app/api/summarize/route.ts"))
    s.write(p / ".env.local", read(".env.local"))
    s.assistant_text("AI 总结已完成,接口已经加了登录保护,只有登录用户才能调用。")
    s.save(logs / "claude" / "projects")
    return "claude-code"

Credential Access

High
Category
Privilege Escalation
Confidence
60% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · tests/helpers/verified.py (reported line 13)May include surrounding context.

python
"E2": {"kind": "code", "path": "supabase/migrations/001.sql", "line": 3,
               "snippet": "alter table notes disable row level security;", "ok": True, "line_corrected": False},
        "E3": {"kind": "probe", "finding_id": "secrets_exposed:.env.local:2", "ok": True, "probe": "secrets_exposed",
               "level": "certain", "path": ".env.local", "line": 2,
               "snippet": "NEXT_PUBLIC_OPENAI_API_KEY=" + OPENAI, "message": "bundled into the browser"},
        "E4": {"kind": "transcript", "session": "s1", "msg_id": "m4", "quote": "暂时先把 key 放前端",
               "ok": True, "ts": "2026-09-13T10:00:00Z", "role": "assistant", "msg_kind": "text"},

Credential Access

High
Category
Privilege Escalation
Confidence
60% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · tests/helpers/verified.py (reported line 33)May include surrounding context.

python
"fix_prompt": "Re-enable row level security on notes and add a policy using (auth.uid() = user_id)."},
        {"id": "P2", "type": "deferred", "question": "open_debts", "plain": "把 OpenAI key 移到服务端",
         "source": ["E4"], "status": "pending", "verdict_plain": "key 还在前端,任何人都能拿走",
         "verdict_dev": "NEXT_PUBLIC_OPENAI_API_KEY in .env.local", "evidence": ["E3"],
         "fix_prompt": "Rename NEXT_PUBLIC_OPENAI_API_KEY to OPENAI_API_KEY and read it only in route.ts."},
        {"id": "P3", "type": "ai_claim", "question": "what_costs", "plain": "接口已经加了登录保护",
         "source": ["E1"], "status": "unknown", "verdict_plain": "没找到能证明这一点的代码",

Credential Access

High
Category
Privilege Escalation
Confidence
60% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · tests/test_probes_stack_secrets_supabase.py (reported line 42)May include surrounding context.

python
def test_client_prefixed_secret_in_env_is_certain(tmp_path):
    write(tmp_path, ".env.local", f"NEXT_PUBLIC_SUPABASE_ANON_KEY={jwt('anon')}\nNEXT_PUBLIC_OPENAI_API_KEY={OPENAI}\n"
                                  "OPENAI_API_KEY_SERVER=fine-on-server\nVITE_ADMIN_SECRET=x\n")
    write(tmp_path, ".env.example", f"NEXT_PUBLIC_OPENAI_API_KEY={OPENAI}\n")
    found = secrets_exposed.run(tmp_path, {"stacks": ["nextjs"]})

Credential Access

High
Category
Privilege Escalation
Confidence
60% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · tests/test_verify_evidence.py (reported line 114)May include surrounding context.

python
def test_masked_or_raw_snippet_matches_a_line_holding_a_secret(tmp_path):
    import redact
    write(tmp_path, ".env.local", f"A=1\nNEXT_PUBLIC_OPENAI_API_KEY={KEY}\n")
    raw = f"NEXT_PUBLIC_OPENAI_API_KEY={KEY}"
    assert verify.check_code({"path": ".env.local", "line": 2, "snippet": raw}, tmp_path)["ok"]
    assert verify.check_code({"path": ".env.local", "line": 2, "snippet": redact.redact(raw)}, tmp_path)["ok"]

Credential Access

High
Category
Privilege Escalation
Confidence
60% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · tests/test_verify_evidence.py (reported line 116)May include surrounding context.

python
import redact
    write(tmp_path, ".env.local", f"A=1\nNEXT_PUBLIC_OPENAI_API_KEY={KEY}\n")
    raw = f"NEXT_PUBLIC_OPENAI_API_KEY={KEY}"
    assert verify.check_code({"path": ".env.local", "line": 2, "snippet": raw}, tmp_path)["ok"]
    assert verify.check_code({"path": ".env.local", "line": 2, "snippet": redact.redact(raw)}, tmp_path)["ok"]

Credential Access

High
Category
Privilege Escalation
Confidence
60% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · tests/test_verify_evidence.py (reported line 117)May include surrounding context.

python
write(tmp_path, ".env.local", f"A=1\nNEXT_PUBLIC_OPENAI_API_KEY={KEY}\n")
    raw = f"NEXT_PUBLIC_OPENAI_API_KEY={KEY}"
    assert verify.check_code({"path": ".env.local", "line": 2, "snippet": raw}, tmp_path)["ok"]
    assert verify.check_code({"path": ".env.local", "line": 2, "snippet": redact.redact(raw)}, tmp_path)["ok"]


def test_trivially_short_snippets_do_not_count_as_evidence(tmp_path):