@promptctl/cc-candybar@1.184.0
Statusline renderer for Claude Code — a JSON5-configurable DSL with daemon-cached data sources, byte-clean palette-aware composition, and OSC8 click verbs.
AI Security Review
scanned 2d ago · by lpm-firewall-aiThe package’s attach path changes Claude Code’s feature gate to enable model-proposed goals, a capability unrelated to its documented statusline purpose. Because this happens during a user-triggered statusline restart and no data theft or destructive behavior was found, this warrants a warning rather than a block.
Decision evidence
public snapshot- The README describes the package as a Claude Code statusline renderer.
- The `propose-goal` seam matches Claude Code’s `tengu_propose_goal` gate and inserts `return true`.
- The source says the model-proposed-goals setting defaults to `auto`, while the server-side feature gate defaults to false.
- Turning the statusline lane on restarts Claude Code under an attach launcher, which resolves and applies the seams.
- The manifest’s lifecycle script is `prepublishOnly`; this behavior is not an install hook.
- The seam runs when the statusline-on action restarts Claude Code, rather than automatically on package install.
- The seam changes an in-memory feature gate; the reviewed code does not show it collecting or sending credentials or other data.
Source & flagged code
10 flaggedSource contains bidi control or invisible Unicode characters associated with Trojan Source attacks.
dist/index.mjsView on unpkg · L10A package entrypoint or install-time lifecycle script reaches a source file with blocking dangerous behavior.
dist/index.mjsView on unpkg · L482Findings
Affected versions and remediation
This report applies to @promptctl/cc-candybar@1.184.0.
See all versions of @promptctl/cc-candybar for other recorded verdicts.
- Review the evidence and your use of @promptctl/cc-candybar@1.184.0 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: .
Think this verdict is wrong? Dispute it on GitHub or read how reports are produced.