npm install -g driftlock

driftlock solves a specific, well-documented problem: AI coding agents frequently exhibit scope creep — modifying files outside the intended change — and lack persistent project memory across sessions.

driftlock acts as a physical guardrail (via a pre-commit hook) and an architectural memory bank (via the context block) to keep agents strictly confined to their authorized scope.

🛡️ Core Protection Features (The Guardrails)

• File & Function Locks: Surgical locking of specific files or even individual AST functions. If an AI agent tries to modify a locked function, the pre-commit hook hard-blocks it.
• Production Path Seals: Permanent locks for critical paths (like billing or auth). These cannot be unlocked normally and require explicit human sign-off via a PR or Jira ticket (driftlock unseal --human-approved=<ticket>).
• Secret Sentinel: A hard-blocking scanner that physically prevents agents from hallucinating and accidentally committing AWS keys, Stripe tokens, or .env leaks.
• Test Coverage Gate: Strict enforcement that blocks any source code changes made by AI agents that aren't accompanied by corresponding tests.
• Dependency Lockdown: Zero-trust dependency management. Automatically locks package.json on init to prevent silent dependency drift and hallucinated packages.

🛠️ Agent Tools (The Tooling)

• Blast Radius Map: Prevent scope creep before it happens. Agents can run driftlock impact <file> to see every file in the codebase that imports a target file before they touch it.
• Rollback Snapshots: Run driftlock save before an agent starts working. If the agent hallucinates or breaks the project, driftlock restore acts as an instant escape hatch to obliterate rogue changes and restore the clean snapshot.
• Context Generation: Agents run driftlock context to generate a token-efficient map of locked vs. active boundaries to feed into their own system prompts.

🤖 The 7 Native AI Skills (Drop-in for Claude/Antigravity/Cline)

Driftlock ships with 7 markdown-based Agent Skills. If your team uses an agent framework, you just point it to these folders and the AI automatically learns how to use Driftlock safely:

• scope-enforcement
• dependency-lockdown
• secret-sentinel
• test-coverage-gate
• rollback-snapshot
• blast-radius
• production-path-lock

Commands & Tiers

Driftlock commands are gated by license tier (Free, Pro, Team).

🟢 Free Tier (No license required)

• driftlock init — Scan repo, generate .driftlock.json
• driftlock lock <file>[:<func>] [reason] — Lock a file or function
• driftlock unlock <file>[:<func>] <reason> — Unlock a target
• driftlock status — Print manifest summary

🟡 Pro Tier (Requires Pro License)

• driftlock save — Auto-snapshot repo state before an agent session
• driftlock restore — Rollback to the last snapshot
• driftlock impact <file> — Show all files that import this file
• driftlock context [task] — Generate AI context block for a task
• driftlock guard [--tests] — Check git diff for violations and secret leaks
• driftlock seal <file> <reason> — Permanent lock for production paths

🔵 Team Tier (Requires Team License)

• driftlock unseal <file> --human-approved=<ticket> <reason> — Release a sealed path
• driftlock trust <file> <reason> — Bypass Secret Sentinel for a specific file (Security Override)
• driftlock godmode <file> — Advanced cross-file refactoring with automated boundary expansion
• driftlock scout — Autonomous repository scanning for architectural drift
• driftlock audit <branch> — AI-driven security and scope-violation auditing for team PRs

License Activation

Get your license here: Driftlock Gumroad

# Activate your license
driftlock login <license-key>

# Check your tier and status
driftlock whoami

# Clear local license cache
driftlock logout

Data Model

All state is stored in .driftlock.json at the root of your repo. The manifest is project state, not a personal config. Commit it so your whole team — and all their AI agents — share the same scope boundaries.