harper-local-dev

Warn

Audited by Gen Agent Trust Hub on Oct 2, 2026

Risk Level: MEDIUMPRIVILEGE_ESCALATIONPERSISTENCECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [PRIVILEGE_ESCALATION]: The documentation in AGENTS.md and rules/running-dev-instances-in-worktrees.md instructs users to execute privileged commands using sudo (e.g., sudo ifconfig lo0 alias 127.0.0.$i up) to configure loopback address aliases on macOS.
  • [PERSISTENCE]: The skill documentation suggests installing a launchd daemon to maintain loopback alias configurations across reboots, which is a persistence mechanism for system configuration changes.
  • [COMMAND_EXECUTION]: The provided scripts/harper-dev.mjs wrapper script utilizes child_process.spawn and execFileSync to execute external binaries, including git and the harper database server.
  • [EXTERNAL_DOWNLOADS]: The documentation provides a link to an external script source on GitHub (https://raw.githubusercontent.com/HarperFast/skills/main/harper-local-dev/scripts/harper-dev.mjs) as an alternative for project setup.
  • [INDIRECT_PROMPT_INJECTION]: The skill exhibits an attack surface by ingesting untrusted data from the local environment.
  • Ingestion points: The script reads the current working directory (process.cwd()), output from git rev-parse, and the contents of a local .env file.
  • Boundary markers: None identified.
  • Capability inventory: The script has capabilities to write files (writeFileSync), delete files (rmSync), and spawn subprocesses (spawn, execFileSync).
  • Sanitization: The script includes a clean function that uses a regular expression (/[^\w.-]+/g) to sanitize project and checkout names before they are used in file paths, which mitigates simple path injection attacks.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Oct 2, 2026, 10:39 PM