Security risk report for @forgecat/fivetaku_fablize v0.1.6
Source Integrity
Low
Static scanner assessed source risk as low, indicating the origin repository (github.com/fivetaku/fablize) is a known, traceable public source.
Author contact and commit hash are explicitly declared, supporting auditability.
No supply-chain red flags such as typosquatted dependencies or unknown binary sources were identified.
Agent Intent
Low
No role-hijacking, credential-reading, or exfiltration instructions are present; all bash commands operate on local plugin state files under ~/.fablize/ and the plugin root.
The profile does not instruct the agent to read sensitive paths (e.g., ~/.ssh, ~/.aws, .env) or POST data to external hosts.
Recommended dependencies (bash, python3, gh) are well-known legitimate tools used for their stated purposes; no typosquatted packages or backdoored templates are embedded.
Details
Evidence
"bash ${FABLIZE_ROOT}/setup/setup.sh" — scoped to plugin root only
"cat ~/.fablize/progress.json 2>/dev/null" — reads only plugin-managed state file
"Optional: gh for the explicit opt-in GitHub star flow" — standard GitHub CLI, opt-in only
Permissions
PRM-003Medium
Four hook scripts (router.sh, gate_prompt.py, gate_stop.py, router-codex.sh) are declared but have no matching tool declarations, making their actual permission scope opaque to static analysis.
The hooks intercept every user prompt submission and stop event (UserPromptSubmit, Stop hooks), giving them broad session-level authority beyond a narrow stated function.
The setup command executes bash scripts that write to CLAUDE.md and ~/.fablize/, which constitutes file-write authority; however, this is scoped to plugin-managed paths and is consistent with the stated setup purpose.
ForgeCat