Back to skill

Security audit

prove-it-works

Security checks for vulnerabilities and agentic risk

Overview

ClawScan rates this skill suspicious because scanners report unsafe command execution and file handling risks, while analysis was partial and could not fully assess some content.

Consider avoiding installation until the command execution and write risks are addressed and the partial analysis is completed. The suspicious verdict indicates meaningful security concerns, but the reported evidence does not establish malicious intent.

Vulnerability Patterns
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • 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
Findings (2)

T09 · Insecure Skill Coding Practices

Warning
Location
scripts/prove.py
Finding

Automatic npm checks execute lifecycle hooks that were not safety-checked

Content
View full analysis

Impact

A hook can run destructive, publishing, or deployment commands that would have been rejected if written directly in the selected script. Such commands run with the checking user's permissions and available credentials.

T09 · Insecure Skill Coding Practices

Warning
Location
scripts/rules.py
Finding

Repository-controlled symlinks redirect rules and ledger appends outside the repository

Content
View full analysis

Impact

An attacker can cause unintended text to be appended to writable external files, corrupting user configuration or other content. The appended content follows the rules or ledger format; unrestricted arbitrary-byte writes or code execution are not established.

Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • System Prompt LeakageDirect Leakage, Indirect Extraction, Tool-Based Exfiltration
Findings (30)

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 · README.md (reported line 32)May include surrounding context.

Optional: run it automatically when Claude Code stops

Claude Code can run a command each time the agent finishes a turn (a Stop hook). This snippet makes the agent keep working while the proof says NOT DONE. It is not installed for you. Add it yourself to .claude/settings.json in your project, or to ~/.claude/settings.json for every project, and adjust the path to where you unzipped the skill:

json
{

Incomplete referenced artifact analysis

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 90)May include surrounding context.

md
- `scripts/diffcheck.py` and `scripts/fakes.py`: the change inspection (causes 5-17), including test commands and CI jobs that cannot fail, tests that cannot fa

Env Variable Harvesting

High
Category
Data Exfiltration
Confidence
70% confidence
Finding

Code enumerates, copies, or searches environment variables for secrets. Bulk environment access can collect credentials unrelated to the skill's stated purpose.

Content

Scanner excerpt · scripts/common.py (reported line 184)May include surrounding context.

python
def _env_secret_values():
    vals = []
    for k, v in os.environ.items():
        if not v or len(v) < 8 or v.lower() in _BORING_VALUES:
            continue
        if not SECRET_ENV_NAME_RE.search(k) or k in ("SSH_AUTH_SOCK", "XDG_SESSION_TYPE", "GPG_AGENT_INFO"):

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 · scripts/common.py (reported line 277)May include surrounding context.

python
def is_env_file(path):
    b = base_of(path).lower()
    if b in (".env", ".envrc", ".dev.vars") or b.startswith(".env.") or b.endswith(".env"):
        return True
    return 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 · scripts/common.py (reported line 277)May include surrounding context.

python
def is_env_file(path):
    b = base_of(path).lower()
    if b in (".env", ".envrc", ".dev.vars") or b.startswith(".env.") or b.endswith(".env"):
        return True
    return False

Env Variable Harvesting

High
Category
Data Exfiltration
Confidence
60% confidence
Finding

Code enumerates, copies, or searches environment variables for secrets. Bulk environment access can collect credentials unrelated to the skill's stated purpose.

Content

Scanner excerpt · scripts/common.py (reported line 345)May include surrounding context.

python
cmd = ["git", "-c", "core.quotepath=off", "-c", "core.fsmonitor=false",
           "-c", "safe.directory=*", "-c", "color.ui=false", "-c", "diff.external=",
           "-c", "core.pager=cat"] + list(args)
    env = dict(os.environ)
    env.update({"GIT_OPTIONAL_LOCKS": "0", "GIT_TERMINAL_PROMPT": "0",
                "GIT_PAGER": "cat", "PAGER": "cat", "LC_ALL": "C"})
    try:

Env Variable Harvesting

High
Category
Data Exfiltration
Confidence
94% confidence
Finding

The entire parent environment, including API tokens and other credentials, is passed to project-controlled commands. In an untrusted repository, tests or build scripts can read those values and transmit them because the offline settings do not enforce network isolation; output redaction does not prevent this access.

Content

Scanner excerpt · scripts/common.py (reported line 493)May include surrounding context.

python
return {"exit": None, "seconds": 0.0, "timed_out": False, "output": "", "error": None, "budget": True}
        if left < timeout:
            timeout = int(max(1, left))
    env = dict(os.environ)
    env.update(OFFLINE_ENV)
    env.update(extra_env or {})
    kw = {}

Direct Prompt Extraction

High
Category
System Prompt Leakage
Confidence
85% confidence
Finding

Skill contains instructions that could directly expose system prompts, internal rules, or hidden instructions to users or external parties.

Content

Scanner excerpt · scripts/diffcheck.py (reported line 409)May include surrounding context.

python
rules.append(("js", None, lambda p: bool(JEST_NAME_RE.match(base_of(p))),
                                  "vitest's default include only runs `*.test.*` / `*.spec.*` files"))
                break
    return rules


def check_test_discovery(change, test_runs, skip_paths=()):

Credential Access

High
Category
Privilege Escalation
Confidence
70% confidence
Finding

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

Content

Scanner excerpt · scripts/diffcheck.py (reported line 598)May include surrounding context.

python
# ---------------------------------------------------------------------------

KEY_FILE_EXT = {".pem", ".key", ".p12", ".pfx", ".jks", ".keystore", ".ppk", ".p8"}
KEY_FILE_NAMES = {"id_rsa", "id_dsa", "id_ecdsa", "id_ed25519", "credentials.json", "service-account.json",
                  ".npmrc", ".pypirc", ".netrc", ".git-credentials"}
GENERIC_SECRET_RE = re.compile(
    r"(?i)\b([\w.-]*(?:secret|passw(?:or)?d|pwd|api[_-]?key|apikey|access[_-]?key|auth[_-]?token|"

Credential Access

High
Category
Privilege Escalation
Confidence
90% confidence
Finding

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

Content

Scanner excerpt · scripts/diffcheck.py (reported line 599)May include surrounding context.

python
KEY_FILE_EXT = {".pem", ".key", ".p12", ".pfx", ".jks", ".keystore", ".ppk", ".p8"}
KEY_FILE_NAMES = {"id_rsa", "id_dsa", "id_ecdsa", "id_ed25519", "credentials.json", "service-account.json",
                  ".npmrc", ".pypirc", ".netrc", ".git-credentials"}
GENERIC_SECRET_RE = re.compile(
    r"(?i)\b([\w.-]*(?:secret|passw(?:or)?d|pwd|api[_-]?key|apikey|access[_-]?key|auth[_-]?token|"
    r"token|private[_-]?key|client[_-]?secret|credential)[\w.-]*)[\"']?\s*(?:[:=]|=>)\s*[\"']([^\"'\s]{10,})[\"']")

Credential Access

High
Category
Privilege Escalation
Confidence
80% confidence
Finding

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

Content

Scanner excerpt · scripts/diffcheck.py (reported line 599)May include surrounding context.

python
KEY_FILE_EXT = {".pem", ".key", ".p12", ".pfx", ".jks", ".keystore", ".ppk", ".p8"}
KEY_FILE_NAMES = {"id_rsa", "id_dsa", "id_ecdsa", "id_ed25519", "credentials.json", "service-account.json",
                  ".npmrc", ".pypirc", ".netrc", ".git-credentials"}
GENERIC_SECRET_RE = re.compile(
    r"(?i)\b([\w.-]*(?:secret|passw(?:or)?d|pwd|api[_-]?key|apikey|access[_-]?key|auth[_-]?token|"
    r"token|private[_-]?key|client[_-]?secret|credential)[\w.-]*)[\"']?\s*(?:[:=]|=>)\s*[\"']([^\"'\s]{10,})[\"']")

Credential Access

High
Category
Privilege Escalation
Confidence
70% confidence
Finding

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

Content

Scanner excerpt · scripts/diffcheck.py (reported line 599)May include surrounding context.

python
KEY_FILE_EXT = {".pem", ".key", ".p12", ".pfx", ".jks", ".keystore", ".ppk", ".p8"}
KEY_FILE_NAMES = {"id_rsa", "id_dsa", "id_ecdsa", "id_ed25519", "credentials.json", "service-account.json",
                  ".npmrc", ".pypirc", ".netrc", ".git-credentials"}
GENERIC_SECRET_RE = re.compile(
    r"(?i)\b([\w.-]*(?:secret|passw(?:or)?d|pwd|api[_-]?key|apikey|access[_-]?key|auth[_-]?token|"
    r"token|private[_-]?key|client[_-]?secret|credential)[\w.-]*)[\"']?\s*(?:[:=]|=>)\s*[\"']([^\"'\s]{10,})[\"']")

The trigger "don't do that again" can turn an ordinary correction into a persistent project rule.

Medium
Category
Trigger Abuse
Confidence
90% confidence
Finding

This phrase is common conversational feedback and does not necessarily request creation of a persistent rule. Restricting it to corrections still leaves ambiguity between feedback about the current response and an instruction to modify .shipsnag/rules.md.

Content

No source excerpt is available for this finding.

Skill declares no tool scope ('permissions' or 'allowed-tools') but code capabilities were detected: env, file_read, file_write, network, shell.

Medium
Category
MCP Least Privilege
Confidence
70% confidence
Finding

Without declared permissions the skill's intent is opaque and cannot be validated.

Content

No source excerpt is available for this finding.

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
75% confidence
Finding

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Content

Scanner excerpt · SKILL.md (reported line 43)May include surrounding context.

md
:** fix each failing check and each problem, then run the proof again, until it passes or what is left needs the person.
   - Fix the code, not the check: do not skip, loosen, delete or re-record a test, add `.only`, add `@ts-ignore` / `# type: ignore` / `eslint-disable`, catch and ignore the error, special-case the test input, stub out the code under test, lower a coverage threshold, add `|| true`, or switch off a CI job. The script flags all of these.
   - **NOT RUN:** do not install anything without asking. Give the person the exact command from the reason column, or ask whether you may run it. If the Stop hook reported it, ask once, then stop.
   - A problem the task really asked for (a dependency the person requested) is not a defect: quote the task sentence that asks for it, in one line, and move on.

4. **Do your own review pass** with `references/review-checklist.md`, even when the verdict is PROVEN. The scripts match names and patterns; they cannot tell whether the change does

The checklist instructs the agent to revert out-of-scope changes without warning that this may discard the user's existing work.

Medium
Category
Output Handling
Confidence
96% confidence
Finding

The checklist tells the agent to revert changes deemed outside the task without first establishing who made them. A diff can include pre-existing user or collaborator work, so following this instruction can discard legitimate edits; placement in a routine verification workflow makes accidental application more likely.

Content

No source excerpt is available for this finding.

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
88% confidence
Finding

Although Git is invoked without a shell and its subcommand is allowlisted, safe.directory=* disables Git's repository-ownership trust protection. Read-only subcommands such as diff can execute configured textconv helpers, and the shown diff invocation disables external diff drivers but not textconv, making inspection of an attacker-configured repository potentially execute code.

Content

Scanner excerpt · scripts/common.py (reported line 349)May include surrounding context.

python
env.update({"GIT_OPTIONAL_LOCKS": "0", "GIT_TERMINAL_PROMPT": "0",
                "GIT_PAGER": "cat", "PAGER": "cat", "LC_ALL": "C"})
    try:
        p = subprocess.run(cmd, cwd=cwd, stdout=subprocess.PIPE, stderr=subprocess.PIPE,
                           timeout=timeout, env=env)
        return p.returncode, p.stdout
    except (OSError, subprocess.TimeoutExpired):

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
94% confidence
Finding

Running project checks is legitimate, but this runner executes project-controlled code with host permissions and optionally a shell, without enforcing the advertised read-only or offline boundaries. The command filters are not a security boundary: permitted interpreters, test hooks, or package scripts can write files or open network connections despite the package-manager environment settings.

Content

Scanner excerpt · scripts/common.py (reported line 505)May include surrounding context.

python
kw["creationflags"] = getattr(subprocess, "CREATE_NEW_PROCESS_GROUP", 0)
    t0 = time.time()
    try:
        p = subprocess.Popen(cmd, cwd=cwd, stdout=subprocess.PIPE, stderr=subprocess.STDOUT,
                             stdin=subprocess.DEVNULL, env=env, shell=shell, **kw)
    except OSError as e:
        return {"exit": None, "seconds": 0.0, "timed_out": False, "output": "", "error": str(e)}

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
70% confidence
Finding

subprocess module calls execute external commands. Without careful input validation, this enables command injection.

Content

Scanner excerpt · scripts/common.py (reported line 539)May include surrounding context.

python
if os.name == "posix":
            os.killpg(p.pid, signal.SIGKILL)
        else:
            subprocess.run(["taskkill", "/F", "/T", "/PID", str(p.pid)],
                           stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL, timeout=20)
    except Exception:
        try:

The automatic-command safety check permits source deletion despite the documented read-only guarantee.

Medium
Category
Excessive Agency
Confidence
99% confidence
Finding

The deletion check accepts ordinary relative targets such as rm -rf src because it rejects traversal and absolute paths rather than requiring a build-output directory. It also explicitly excludes rimraf from this validation, so an automatically detected check can pass this gate while deleting source files, contrary to the read-only claim.

Content

No source excerpt is available for this finding.

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
75% confidence
Finding

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Content

Scanner excerpt · scripts/diffcheck.py (reported line 208)May include surrounding context.

python
6, "Tests weakened in %s" % p,
                "In %s the change leaves the tests weaker than before: %s." % (p, "; ".join(reasons)),
                ev[:6],
                "The suite now passes without checking what it used to check, so a green run no longer proves the behaviour works.",
                ["Restore the removed or loosened assertions and un-skip the tests, then make the code pass them. "
                 "If an expectation really changed, assert the new exact value instead of a looser one."],
                file=p, line=(skips or a_commented or [(None, "")])[0][0], tier=tier, also=[5] if skips else []))

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
75% confidence
Finding

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Content

Scanner excerpt · scripts/diffcheck.py (reported line 263)May include surrounding context.

python
(", and the new snapshot contains %s" % ", ".join("`%s`" % b for b in bad)) if bad else
                ("" if (mentions or expl) else ", and nothing in the task asks for different output")),
            [_ev(f.path, ln, t) for ln, t in f.added[:3]] or ["%s (deleted)" % f.path],
            "Updating snapshots to whatever the code now prints turns a failing test into a passing one without checking the output is right.",
            ["Diff the old and new snapshot by eye. If the new output is wrong, restore the snapshot and fix the code; "
             "only re-record when the task asked for the new output."],
            file=f.path, tier=tier))

MCP server referenced without pinned version: 'npx prisma'.

Medium
Category
MCP Rug Pull
Confidence
70% confidence
Finding

npx commands without a version suffix (e.g. @1.0.0) create a rug-pull risk if the upstream server is compromised and publishes a malicious update.

Content

No source excerpt is available for this finding.

MCP server referenced without pinned version: 'npx drizzle-kit'.

Medium
Category
MCP Rug Pull
Confidence
70% confidence
Finding

npx commands without a version suffix (e.g. @1.0.0) create a rug-pull risk if the upstream server is compromised and publishes a malicious update.

Content

No source excerpt is available for this finding.

Tainted flow: 'path' from sys.stdin.read (line 1754, user input) → open (file write)

Medium
Category
Data Flow
Confidence
65% confidence
Finding

Data from a source is assigned to a variable that is later passed to a sink, creating a variable-mediated taint flow.

Content

Scanner excerpt · scripts/prove.py (reported line 1388)May include surrounding context.

python
def _save_memo(path, data):
    try:
        with open(path, "w") as fh:
            json.dump(data, fh)
    except OSError:
        pass