@markjaquith/agency@3.8.0
Manage agentic work across repositories with durable workbases
AI Security Review
scanned 14d ago · by lpm-firewall-aiThe postinstall hook installs a package-owned extension that modifies Pi’s agent context, a warning-level lifecycle risk. The inspected source did not establish a concrete malicious chain or credential exfiltration.
Decision evidence
public snapshot- package.json runs the Pi extension installer during postinstall and its uninstaller during preuninstall.
- The installer copies the package-owned extension into the user’s Pi agent extensions directory.
- The extension adds workbase instructions and context to Pi’s system prompt when a workbase is found.
- IntegrationService also manages project integrations under .agency and .opencode paths.
- The extension obtains context by running the agency CLI for the discovered workbase.
- The inspected extension behavior reads workbase instructions and registers Pi event handlers; no credential forwarding or unrelated network recipient was identified.
- The CLI integration command exposes explicit status and sync operations.
Source & flagged code
10 flagged · loading sourcePackage defines install-time lifecycle scripts.
package.jsonView on unpkgInstall-time lifecycle script is not statically allowlisted and needs review.
package.jsonView on unpkgPackage source references dynamic require/import behavior.
src/utils/interactive-loader.tsView on unpkg · L3Findings
Affected versions and remediation
This report applies to @markjaquith/agency@3.8.0.
See all versions of @markjaquith/agency for other recorded verdicts.
- Review the evidence and your use of @markjaquith/agency@3.8.0 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.