Autonomy Report by sixtysecondsapp/use60

Available Context & Tools

@_platform-references/org-variables.md @_platform-references/capabilities.md

Instructions

You are executing the Autonomy Report skill. Your job is to query the user's autopilot confidence data and produce a clear, motivating summary of their current automation profile — what's running on autopilot, what's nearly there, how much time is being saved, and what they should promote next.

Data Gathering

Fetch the user's autopilot confidence data by querying the autopilot_confidence table:

execute_action("get_autopilot_confidence", { user_id: "${user_id}" })

This returns rows from the autopilot_confidence table. Each row has:

action_type: string identifier (e.g. crm.note_add, email.send, task.create)
current_tier: one of disabled, suggest, approve, auto
score: 0–100 confidence score (higher = more trusted)
approval_rate: fraction of executions approved by the user (null if no data)
clean_approval_rate: fraction approved without edits (null if no data)
edit_rate: fraction where user edited before approving (null if no data)
rejection_rate: fraction rejected (null if no data)
undo_rate: fraction undone after auto-execution (null if no data)
total_signals: total number of interactions recorded
total_approved: count of approvals
total_rejected: count of rejections
total_undone: count of undos
last_30_score: score over the last 30 days (null if insufficient data)
days_active: days since first signal
promotion_eligible: boolean — true if ready for next tier
cooldown_until: ISO timestamp if in cooldown after a rejection/undo (null if not)
never_promote: boolean — user has locked this action out of auto promotion
extra_required_signals: additional signals needed before promotion eligibility
first_signal_at: ISO timestamp of first signal
last_signal_at: ISO timestamp of most recent signal

If no rows are returned, the user has not interacted with autopilot yet — present an onboarding message.

Computed Metrics

Calculate these aggregates from the returned rows before writing the report:

Autonomy Score (0–100):

autonomy_score = (count of rows where current_tier = 'auto') / (total rows) * 100
Round to nearest integer.

Tier Counts:

auto_count: rows where current_tier = 'auto'
approve_count: rows where current_tier = 'approve'
suggest_count: rows where current_tier = 'suggest'

Time Saved Per Week (hours): Use these per-action time estimates (in seconds) to compute weekly savings:

Action Type	Seconds Saved
`crm.note_add`	120
`crm.activity_log`	30
`crm.contact_enrich`	300
`crm.next_steps_update`	60
`crm.deal_field_update`	45
`crm.deal_stage_change`	30
`crm.deal_amount_change`	30
`crm.deal_close_date_change`	30
`email.draft_save`	0
`email.send`	600
`email.follow_up_send`	480
`email.check_in_send`	300
`task.create`	60
`task.assign`	30
`analysis.risk_assessment`	900
`analysis.coaching_feedback`	1200
(all other action types)	60

For each row at auto or approve tier:

signals_per_week = (total_signals / 90) * 7
multiplier = 1.0 for auto, 0.7 for approve (HITL reduces savings)
time_saved_seconds += signals_per_week * action_time_seconds * multiplier

time_saved_hours_week = time_saved_seconds / 3600

Promotion Candidates: Rows that are NOT at auto and NOT never_promote = true, sorted by:

promotion_eligible = true first
Then highest score
Then lowest extra_required_signals

Take the top 3 candidates.

Next-Tier Labels:

suggest -> next tier is approve
approve -> next tier is auto

Output Format

Render a rich, plain-text formatted report using block characters and markdown-compatible structure. Do NOT use emoji — use action_type names and Lucide-style labels instead.

Report Template

AUTONOMY REPORT
───────────────────────────────────────

OVERALL AUTONOMY SCORE
[render a 20-char block bar where filled blocks (█) represent score, empty (░) represent remainder]
[score]% autonomous   [auto_count] auto / [approve_count] approve / [suggest_count] suggest

[One-line interpretation:]
- 0%: "Autopilot is just getting started — every approval builds trust."
- 1–25%: "Early stage — the AI is learning your patterns."
- 26–50%: "Growing autonomy — several actions are now fully automated."
- 51–75%: "Strong autopilot — more than half your tracked actions run hands-free."
- 76–99%: "Near-full autonomy — you're running lean."
- 100%: "Maximum autonomy — fully hands-free across all tracked actions."

TIME SAVED THIS WEEK
~[X.X] hrs saved      [X] actions ran on autopilot (last 30 days)

───────────────────────────────────────

AUTO (fully hands-free)
[For each auto-tier row:]
  [action_type label]   score: [score]/100   [approval_rate*100]% approval rate
  [if undo_rate > 0.05]: "  Note: [undo_rate*100]% undo rate — watch this one"

APPROVE (you review, then it runs)
[For each approve-tier row:]
  [action_type label]   score: [score]/100   [total_approved] approved / [total_rejected] rejected
  [if promotion_eligible]: "  Ready to promote to auto"

SUGGEST (recommendation only)
[For each suggest-tier row:]
  [action_type label]   score: [score]/100   [total_signals] signals recorded

───────────────────────────────────────

WHAT'S NEXT
[Show top 3 promotion candidates:]

[For each promotion_eligible = true candidate:]
  [action_type label] is ready for promotion
  Current tier: [current_tier] -> [next_tier]   Score: [score]/100
  "[action_type] has [total_approved]/30 clean approvals. Want me to upgrade it to [next_tier]?"

[For each candidate NOT yet promotion_eligible:]
  [action_type label] needs [extra_required_signals] more signals
  Current tier: [current_tier]   Score: [score]/100   [total_signals] signals so far

[If no candidates:]
  All tracked actions are either at 'auto' or locked from promotion.

───────────────────────────────────────

RECENT ACTIVITY
[List the 3 most recently active rows by last_signal_at:]
  [action_type label]   last seen: [relative time, e.g. "2 days ago"]   tier: [current_tier]

Action Type Labels

Map raw action_type values to human-readable labels:

action_type	Label
`crm.note_add`	CRM Note
`crm.activity_log`	Activity Log
`crm.contact_enrich`	Contact Enrichment
`crm.next_steps_update`	Next Steps Update
`crm.deal_field_update`	Deal Field Update
`crm.deal_stage_change`	Deal Stage Change
`crm.deal_amount_change`	Deal Amount Update
`crm.deal_close_date_change`	Close Date Update
`email.draft_save`	Email Draft Save
`email.send`	Email Send
`email.follow_up_send`	Follow-Up Email
`email.check_in_send`	Check-In Email
`task.create`	Task Creation
`task.assign`	Task Assignment
`analysis.risk_assessment`	Risk Assessment
`analysis.coaching_feedback`	Coaching Feedback
(unknown)	use the raw `action_type` value, replacing dots with spaces

Block Bar Rendering

For a 20-character progress bar at score S:

filled = round(S / 100 * 20)
empty = 20 - filled
Bar: "█".repeat(filled) + "░".repeat(empty)

Example at 65%: █████████████░░░░░░░

Call-to-Action Phrasing

For each promotion_eligible = true candidate, produce a specific call to action:

If current_tier = 'approve' and next_tier = 'auto':

"Want me to upgrade [Label] to auto? I've got [total_approved]/[total_approved + total_rejected] clean approvals — that's [approval_rate*100 rounded]%."
If current_tier = 'suggest' and next_tier = 'approve':

"[Label] is ready to move to approve mode. Should I enable it? You'll get a preview before each action runs."

Error Handling

No data returned (new user)

"Autopilot is warming up. As you interact with the platform, I'll track which actions I can handle for you and start earning trust tier by tier. Check back after a few days of activity."

All actions at 'auto' (100% autonomy)

"You've reached maximum autonomy — every tracked action type is running hands-free. Nothing to promote. I'll continue monitoring for reliability and flag anything that needs attention."

All actions locked (`never_promote = true` on all non-auto rows)

"All remaining actions are locked from promotion by your preferences. If you'd like to re-enable promotion for any action, visit Autopilot Settings."

Partial data (some rows missing `approval_rate`)

Skip the approval_rate display for those rows. Do not show null or "N/A" — simply omit the metric.

Quality Checklist

Before returning:

Autonomy score computed correctly (auto count / total count)
Block bar matches score visually
Time saved uses the correct per-action seconds table and weekly extrapolation
Promotion candidates sorted correctly (eligible first, then by score)
Call-to-action phrasing includes specific signal counts, not generic text
Action type labels are human-readable (not raw crm.note_add style)
No emoji used anywhere in the output
Undo rate warnings included where undo_rate > 0.05
Cooldown actions noted if cooldown_until is in the future

Autonomy Report