recovery-readiness

name: recovery-readiness description: Produce a bounded RecoveryProposal for today's session by consuming the runtime-computed classified_state + policy_result and applying judgment-only steps — action matrix, rationale prose, and vendor cross-check. The runtime already did every band, every score, and every policy rule; this skill does not re-derive them. allowed-tools: Read, Bash(hai state snapshot *), Bash(hai state read *), Bash(hai clean *), Bash(hai propose *), Bash(hai intent list *), Bash(hai target list *) disable-model-invocation: false

Recovery Readiness

All arithmetic happens in code. Your job is: read the bundle, honour the policy result, pick an action, write the rationale, persist.

Load the bundle

hai state snapshot --as-of <today> --user-id <u> --evidence-json <hai clean output>

Under snapshot.recovery you receive five blocks:

• evidence — cleaned inputs (sleep_hours, resting_hr, hrv_ms, soreness_self_report, planned_session_type, active_goal, …).
• raw_summary — deltas / ratios / coverage fractions. Context only; never re-derive.
• classified_state — sleep_debt_band, resting_hr_band, hrv_band, training_load_band, soreness_band, coverage_band, recovery_status, readiness_score, uncertainty. Source of truth.
• policy_result — policy_decisions[], forced_action, forced_action_detail, capped_confidence, and (v0.1.14 W-PROV-1) optional evidence_locators[] when an R-rule that emits source-row provenance fired (today: R6 resting_hr_spike_3_days_running). Source of truth.
• missingness — per state_model_v1.md §5.

Protocol

1. If the policy forced an action, use it

If policy_result.forced_action is set, action is that value and action_detail is policy_result.forced_action_detail. Confidence: low for defer_decision_insufficient_signal, else moderate. Skip the action matrix; jump to rationale.

2. Otherwise, pick from the action matrix

Keyed on classified_state.recovery_status × evidence.planned_session_type:

3. Confidence

Default from classified_state.coverage_band: full → high, partial → moderate, sparse → moderate, insufficient → low. If policy_result.capped_confidence is set, it lowers the default but never raises it. The vendor cross-check below may lower further.

4. Vendor cross-check

The snapshot also carries Garmin's own signals under recovery.today:

• training_readiness_level — vendor-authored categorical band (e.g. "High", "LOW"). Pair it with training_readiness_component_mean_pct, which is a locally-computed arithmetic mean of five Garmin component pcts (training_readiness_sleep_pct, _hrv_pct, _stress_pct, _sleep_history_pct, _load_pct) and can disagree with the vendor level because Garmin applies internal weighting.
• all_day_stress, body_battery_end_of_day — passive aggregates. Informational only in rationale.
• garmin_acwr_ratio + acwr_status — vendor load band.
• moderate_intensity_min, vigorous_intensity_min, total_distance_m — running activity for the day.

If training_readiness_level disagrees with classified_state.recovery_status, cap confidence at moderate and add agent_vendor_readiness_disagreement to uncertainty. If the local mean and the categorical level diverge, add training_readiness_weighting_disagreement — do not trust the local mean alone.

5. Rationale (5–8 lines)

One line per band that informed the decision. Name the band; do not re-derive it.

Examples: sleep_debt=<band>, resting_hr_vs_baseline=<band>, hrv_vs_baseline=<band>, soreness_signal=<band>, training_load=<band>, body_battery_end_of_day=<int> (informational), active_goal=<goal> (if present), agent_vendor_readiness_disagreement (if applicable).

6. Uncertainty

Start with classified_state.uncertainty (already sorted + deduped). Append any tokens you added (e.g. *_unavailable_at_source derived from the snapshot's missingness token, agent_vendor_readiness_disagreement, training_readiness_weighting_disagreement). Re-sort alphabetically; deduplicate.

7. Stop at the proposal boundary

This skill does not schedule follow-up reviews. follow_up, review_at, review_question, review_event_id, recommendation_id, and daily_plan_id are synthesis/daily-owned surfaces and are forbidden in proposal payloads. The runtime attaches review scheduling only after hai synthesize / hai daily commits a final recommendation.

Output

Emit a RecoveryProposal JSON and call hai propose --domain recovery --proposal-json <path> --base-dir <root>. The propose tool validates the shape and appends to proposal_log; it is your determinism check.

proposal_id = prop_<for_date>_<user_id>_recovery_01 on first write; revisions get _02, _03. Use hai propose --replace when revising the same day's proposal with new skill output — the runtime creates a new revision leaf and forward-links the prior one. Identical-payload replay under --replace is a no-op.

Copy snapshot.recovery.policy_result.policy_decisions into the output's policy_decisions verbatim. v0.1.14 W-PROV-1: if snapshot.recovery.policy_result.evidence_locators is present (R6 spike firing), copy that list verbatim into the proposal's evidence_locators field — do NOT derive locators yourself; the runtime computed them. If the field is absent, omit evidence_locators from the proposal. Synthesis owns downstream: daily_plan_id, recommendation_id, X-rule-applied action_detail, review scheduling, and locator carry-forward to recommendation_log.evidence_locators_json. This skill emits one domain's bounded proposal and stops.

Invariants

• You never compute a band, a score, or a ratio. classified_state is the source of truth.
• You never evaluate a policy rule (R1 – R6). policy_result is the source of truth; you honour forced_action and capped_confidence.
• You never apply X-rule mutations. Synthesis owns that in Phase 2; this skill emits one domain's bounded proposal.
• You never emit an action outside the v1 enum. hai propose enforces this.
• You never fabricate values for missing evidence; missing stays missing.
• If a decision isn't reasoned in rationale[] or policy_decisions[], it didn't happen.