Caveat verdict

gstack

88
🟢 Trusted
No high-risk patterns surfaced by the deep scan — automated capability review, not behavioral proof.

Headless browser QA testing skill for site verification; uses agent_memory for session tracking as part of its stated browser testing purpose with no suspicious exfiltration.

⚠ Flagged for review — coarse, uncorroborated signal, not a confirmed exploit. Review the config yourself before installing.

Automated static analysis — not a human review. Caveat flags capabilities, not confirmed intent, and can produce false positives. Disagree with this verdict? Use Dispute below.

0
security
90
transparency
100
maintenance

Permission integrity

Accesses agent memory/configuration files

agent_memory

Findings (19)

Pattern match high

Writes to SKILL.md — self-modifying skill

SKILL.md · prose · downgraded · SKILL.md` and follow the "Inline upgrade flow" (auto-upgrade if configured, othe

Pattern match high

Pipe-to-shell pattern (curl | sh) — supply chain attack vector

SKILL.md · prose · downgraded · curl -fsSL https://bun.sh/install | bash

Pattern match high

Pipe to bash — executes piped content as shell commands

SKILL.md · prose · downgraded · | bash

Pattern match high

<script> tag in markdown — potential code injection

browse/src/cookie-picker-ui.ts · prose · downgraded · <script>

Pattern match high

Accesses sensitive system files

browse/test/commands.test.ts · prose · downgraded · /etc/passwd

Pattern match high

Possible hardcoded credential

browse/test/cookie-import-browser.test.ts · prose · downgraded · password = "test-keychain-password

Pattern match high

Recursive delete from root or home — destructive command

README.md · prose · downgraded · rm -rf ~

Pattern match medium

References agent configuration files

document-release/SKILL.md · frontmatter · CLAUDE.md

Pattern match medium

Accesses system credential store

ARCHITECTURE.md · prose · downgraded · Keychain

Pattern match medium

Instructs covert action — may act without user awareness

browse/src/write-commands.ts · prose · downgraded · silently

Pattern match medium

Uses spawn() — can execute external programs

browse/src/cli.ts · prose · downgraded · spawn(

Pattern match medium

Bun.spawn — spawns subprocess in Bun runtime

browse/src/cli.ts · prose · downgraded · Bun.spawn

Pattern match medium

HTTP request to bare IP address — common in malicious payloads

browse/src/cli.ts · prose · downgraded · http://127.0.0.1

Pattern match medium

Dynamic import() — loads module at runtime

browse/src/meta-commands.ts · prose · downgraded · import('

Pattern match medium

References child_process — can spawn system processes

browse/test/commands.test.ts · prose · downgraded · child_process

Pattern match medium

Uses exec() — may execute shell commands

test/skill-validation.test.ts · prose · downgraded · exec(

Pattern match low

Opens WebSocket connection

ARCHITECTURE.md · prose · downgraded · WebSocket

Pattern match low

Bun file read API

browse/src/server.ts · prose · downgraded · Bun.file(

Pattern match low

Accesses sensitive environment variables

test/skill-e2e.test.ts · prose · downgraded · process.env.ANTHROPIC_API_KEY

Why the tier is capped

Execution sink present in raw bytes (Hard Floor: class A/B/D). Final tier capped at Caution — cannot be lifted by any downgrade, example-payload opt-in, or allowlist.

Permissions & capabilities

No declared permissions — minimal attack surface.

agent_memory

Is this flag fair?

Check another skill Browse the registry Auditing your own skills or configs? Use the API