Self Attempt 01

Asteria Investment Office Portfolio Task Skill

Purpose

Solve institutional portfolio management tasks that require querying the Asteria shared environment API and producing strict JSON output matching a provided schema template.

Prerequisites

• The task provides input/prompt.txt, input/payloads/<request>.json, and input/payloads/answer_template.json
• Environment URL is fixed: GDPEVO_ENV_BASE_URL=http://34.46.77.124:8010 (or as specified in environment_access.md)
• Do not read env/, test/, or eval/ directories

Step-by-Step Procedure

1. Parse the Request

Read input/prompt.txt and input/payloads/*.json to identify:

• Target portfolio ID (e.g., PF-EN-ALTA, PF-FI-LUMEN, PF-MA-HELIO)
• Task type: trade strategy, rebalance/rotation, allocation review, correlation analysis, committee decision file
• Stale data warnings in the payload — these are critical; always prefer current API data
• Required output keys from answer_template.json

2. Query the Shared Environment API

Use the API in this order to gather current data. Always prefer current API records over any stale payload data.

Discovery:

curl -s http://34.46.77.124:8010/api/catalog

Returns all valid portfolio_ids, policy_ids, index_ids, bond_instrument_ids, issuer_ids, and opportunity_sets.

Portfolio-specific:

curl -s http://34.46.77.124:8010/api/portfolios/<portfolio_id>

Returns as_of_date, holdings (with instrument_id, quantity_usd_m, sleeve, notes), constraints, market_value_usd_m.

Constraints & Policies:

curl -s http://34.46.77.124:8010/api/policies

Fetch the portfolio's constraint_policy_id to get duration_band_years, max_hy_allocation_pct, target_hy_reduction_pct, correlation_high_threshold, etc.

Bond Universe:

curl -s http://34.46.77.124:8010/api/instruments/bonds

Returns all bonds with rating_bucket (IG/HY), modified_duration_years, yield_to_maturity_pct, sector, subsector, energy_linked, candidate (true/false), recommended_theme_tags, issuer_id.

Issuers / Watchlist:

curl -s http://34.46.77.124:8010/api/issuers

Returns watchlist flags and issuer research. Use this to identify WATCHLIST_RISK instruments.

Energy Market Signals (for energy portfolios):`

curl -s http://34.46.77.124:8010/api/market/energy

Returns pitch_themes, commodity signals, and a stale_data_warning date threshold.

Equity Index Data (for correlation/allocation tasks):`

curl -s http://34.46.77.124:8010/api/indices
curl -s http://34.46.77.124:8010/api/index-levels/<index_id>

Index levels are monthly. Compute simple returns as (level_t / level_{t-1}) - 1, then Pearson correlation.

Allocation Views & Macro Signals:

curl -s http://34.46.77.124:8010/api/allocation/opportunity-sets
curl -s http://34.46.77.124:8010/api/allocation/prior-views
curl -s http://34.46.77.124:8010/api/macro-signals

Prior views show view (OW/N/UW), conviction, and previous_quarter. Macro signals show score and rationale_code.

3. Resolve Data Precedence

If the payload contains a stale_local_note or stale_data_warning, always refresh from the API before computing:

• Use the portfolio's current as_of_date from /api/portfolios/<id> as the as_of_date in the answer.
• Use current holdings quantities, not payload quantities.
• Use current bond ratings, durations, and yields from /api/instruments/bonds.
• Use current macro signals and prior views from /api/allocation/* and /api/macro-signals.

4. Perform Calculations

Correlation (for equity tasks):

1. Fetch 12 monthly levels for each requested index_id.
2. Compute simple monthly returns: r_t = (level_t / level_{t-1}) - 1.
3. Compute Pearson correlation across the return series.
4. Round to 3 decimal places.
5. Sort index IDs alphabetically within each pair.
6. Identify:
   • highest_concentration: pair with highest correlation
   • best_diversifier: pair with lowest correlation

Post-Trade Metrics (for credit tasks):

1. Start with current holdings from /api/portfolios/<id>.
2. Apply requested trades (SELL reduces quantity, BUY adds).
3. Compute new market_value_usd_m using bond prices (treat quantity_usd_m as market value).
4. Compute hy_allocation_pct = sum of HY holdings / total market value * 100.
5. Compute weighted duration and weighted YTM:
   • weighted_dur = sum(holding_qty * bond_duration) / total_mv
   • weighted_ytm = sum(holding_qty * bond_ytm) / total_mv
6. Round to 2 decimal places unless template specifies otherwise.

HY Reduction:

• hy_reduction_pct_points = pre_trade_hy_pct - post_trade_hy_pct
• Compare against policy target_hy_reduction_pct or payload minimum_preferred_hy_reduction_pct_points.

Watchlist Handling:

• Query /api/issuers or check bond recommended_theme_tags for WATCHLIST_RISK.
• Sell all watchlist holdings if the task requires clearing watchlist exposure.
• Never buy a candidate bond whose issuer or tags indicate watchlist risk if avoid_new_watchlist_buy: true.

Allocation Views (for multi-asset/policy tasks):

• Map each opportunity_set to its asset_class using /api/allocation/opportunity-sets.
• Get prior_view from /api/allocation/prior-views.
• Get signal_score from /api/macro-signals.
• Derive view and change (UP/DOWN/UNCHANGED) by comparing current macro signal direction vs. prior view.
• Select rationale_code from the allowed enum matching the macro signal's rationale_code.
• Set conviction based on signal magnitude (e.g., |score| > 0.4 → HIGH, 0.2–0.4 → MEDIUM, < 0.2 → LOW).

5. Populate the Answer Template

• Include all required top-level keys exactly as listed in answer_template.json.
• Respect enum allowed_values exactly; do not invent values.
• Respect ordering rules (e.g., SELL before BUY, then alphabetical; or specific opportunity_set order).
• Use correct precision (decimal places) for each numeric field.
• Set data_precedence to current_environment_over_stale_payload whenever stale notes exist.

6. Validate Before Returning

• Verify JSON is parseable and contains no markdown fences or narrative text.
• Check that all required / required_top_level_keys are present.
• Check enum values against allowed lists.
• Check numeric precision matches template.
• Verify ordering rules (alphabetical, SELL-before-BUY, or payload-defined sequence).
• Confirm as_of_date matches the current API portfolio record, not the payload date.

Common Pitfalls

• Using stale payload data: If the payload has stale_local_note, override with API data.
• Wrong as_of_date: Use the API portfolio as_of_date, not the request's memo date.
• Missing required keys: The template explicitly lists required keys; verify every one.
• Incorrect enum values: Rationale codes, action types, and view codes must match exactly (e.g., GROWTH_IMPROVES, not growth_improves).
• Precision errors: Round exactly as specified (usually 1, 2, or 3 decimal places).
• Watchlist oversight: Always cross-check /api/issuers and bond recommended_theme_tags for WATCHLIST_RISK.
• Correlation pair ordering: Index IDs within each pair must be sorted alphabetically.

Example API Session

BASE="http://34.46.77.124:8010"
PORTFOLIO="PF-FI-LUMEN"

# 1. Catalog
curl -s "$BASE/api/catalog" | jq .

# 2. Portfolio holdings & constraints
curl -s "$BASE/api/portfolios/$PORTFOLIO" | jq .

# 3. Bond universe for candidate selection
curl -s "$BASE/api/instruments/bonds" | jq '.[] | select(.candidate==true)'

# 4. Issuer watchlist status
curl -s "$BASE/api/issuers" | jq '.[] | select(.watchlist==true)'

# 5. Index levels for correlation
curl -s "$BASE/api/index-levels/IDX_EM" | jq '.levels'