Clockwork Compliance Check

Use this skill to evaluate timer compliance for a team using clockwork mcp (not direct REST endpoints).

Working windows (Asia/Ho_Chi_Minh):

• Morning: 08:30-12:00
• Afternoon: 13:30-17:00

Trigger Phrases

• "run morning timer compliance check"
• "check who has not started timer"
• "check who has not resumed after lunch"
• "find idle gaps longer than 75 minutes"
• "run end of day timer compliance"

Tools

• clockwork mcp:
  • get_all_active_timers (required, primary for team-wide timer state)
  • get_active_timers (optional fallback)
  • get_worklogs (required)
  • search_issues (optional)
• Local shell tools for roster intake:
  • curl (download CSV roster from URL)
  • python or awk (CSV parse/validation)
• Calendar MCP (built-in Manus): out-of-office and leave exclusion
• Jira MCP (sooperset/mcp-atlassian, optional): enrichment and escalation context

Preconditions

• clockwork mcp is configured and reachable.
• Team roster is fetched from a CSV URL (or provided directly by user).
• Timezone is fixed to Asia/Ho_Chi_Minh.
• Compliance policy is Hybrid:
  • Total logged time target per day: default 7.5 hours
  • Idle gap warning threshold: default 75 minutes

Guardrails

• Ask for csv_url if not present in prompt.
• Validate roster schema before running compliance checks.
• Auto-detect CSV delimiter (; or ,); Atlassian exports commonly use ;.
• Accept Atlassian header aliases:
  • User id -> accountId
  • User name -> displayName
  • User status -> active
• Exclude users with approved leave/out-of-office in the checked window.
• Do not evaluate outside working windows.
• Use get_all_active_timers once per run and compare against roster to find missing users.
• Mark user as unknown when required source data is missing.
• Return machine-readable output for downstream alert routing.

Workflow Router

1. Load references/workflows/compliance-check-workflow.md.
2. Output records are the required input for clockwork-alert-router.

Output Contract

Return JSON-ready records with these fields:

• date
• timezone
• run_type
• member: displayName, accountId, email
• status: compliant | warning | high | unknown
• violations: missing_start | missing_resume | hours_below_target | long_idle_gap
• metrics: total_logged_hours, longest_idle_gap_minutes, first_timer_at, last_timer_at
• exclusion_reason
• recommended_action: none | email | jira_escalation
• roster_source: csv_url
• invalid_roster_rows
• timer_snapshot_meta: total_accounts, total_timers, cached_at

Failure Handling

• Missing csv_url: ask user and stop.
• CSV download/parse failure: return roster_fetch_error or invalid_roster_schema.
• get_all_active_timers failure: retry once, fallback to per-user get_active_timers only if fallback is explicitly allowed for this run.
• clockwork mcp global failure: return partial result with clockwork_mcp_error.
• Calendar unavailable: continue and mark calendar_unavailable=true.