loop

/anty:loop — Persistent Operation

/anty:loop — Persistent Operation

Trigger

User runs /anty:loop [interval]

Arguments

  • /anty:loop 30m — Run scan every 30 minutes
  • /anty:loop 1h — Run scan every hour
  • /anty:loop 2h — Run scan every 2 hours
  • /anty:loop stop — Stop the running loop

Behavior

Establish a persistent scan loop that repeatedly executes /anty:scan at the specified interval. Uses the loop skill pattern for continuous operation within a Claude Code session.

Starting the Loop

  1. Parse the interval argument (e.g., 30m, 1h, 2h)
  2. Validate interval is between 10m and 24h
  3. Display confirmation: 
    Anty loop started.
Interval: [interval]
Next scan: [timestamp]
Use /anty:loop stop to end.
  4. Begin the loop cycle

Loop Cycle

Each iteration:

  1. Read state: Load all .anty/ files to get current state
  2. Execute scan: Run the full /anty:scan 10-point analysis
  3. Evaluate findings:
    • If issues found with severity >= warning:
      • Generate recovery Action proposals
      • Notify user with a summary of findings
    • If no issues: log clean scan silently
  4. Save results: Write scan results to .anty/history/scans/[date].yaml
  5. Append to scratchpad: Add scan summary to .anty/scratchpad.md
  6. Wait: Sleep for the specified interval
  7. Repeat: Go to step 1

Notification Thresholds

Only notify the user (interrupt their work) when:

  • Buffer zone transitions (GREEN->YELLOW, YELLOW->RED)
  • Pipeline deal stalls detected (2x average dwell time)
  • Failure pattern detected (3+ consecutive failures)
  • Anti-pattern detected
  • Growth rate drops below 5%/week
  • Environmental disruption risk detected
  • Oscillation pattern detected

For non-critical findings, accumulate and report at next user interaction.

Notification Format

----------------------------------------------------
ANTY LOOP — [timestamp]
----------------------------------------------------
[severity] [check name]: [one-line summary]

[n] recovery proposals generated.
Run /anty:actions to review.
----------------------------------------------------

Stopping the Loop

When user runs /anty:loop stop:

  1. End the loop cycle
  2. Display summary of the session: 
    Anty loop stopped.
Session duration: [duration]
Scans completed: [n]
Issues detected: [n]
Recovery proposals generated: [n]
  3. Save final state to .anty/

Loop Implementation

Use the standard loop skill pattern:

  • Leverage Claude Code's ability to run persistent background operations
  • Use bash sleep between iterations with the specified interval
  • Track loop state (running/stopped, iteration count, start time)
  • Handle graceful shutdown on /anty:loop stop

Interval Parsing

  • Nm — N minutes (e.g., 30m = 1800 seconds)
  • Nh — N hours (e.g., 1h = 3600 seconds)
  • Minimum: 10m
  • Maximum: 24h
  • Default (no argument): 1h

Error Handling

  • If a scan fails, log the error and continue to next iteration
  • If .anty/ files are corrupted, attempt repair and notify user
  • If session is about to end, save all state before termination