HK IPO Decision Engine#

Use this skill for action-oriented decisions.

Runtime#

This publish bundle includes the required CLI runtime under runtime/hkipo-next.

From the skill folder:

bash

cd <skill_dir>
uv run --directory runtime/hkipo-next hkipo-next ...

By default these commands persist review records in ~/.hkipo-next/data/hkipo.db.

Preconditions#

• Prefer to have an active parameter version.
• Prefer to have a usable profile or explicit CLI overrides for risk, budget, and financing preference.
• If the user only wants rough market discovery, do not jump straight to this skill.

Choose the Right Command#

• Use score for factor-level explanation.
• Use decision-card for a concise execution plan.
• Use batch for multiple symbols or the saved watchlist.

Commands#

Score one symbol:

bash

cd <skill_dir>
uv run --directory runtime/hkipo-next hkipo-next score 2476 --format json

Score with explicit scenario overrides:

bash

cd <skill_dir>
uv run --directory runtime/hkipo-next hkipo-next score 2476 \
  --parameter-version v0003 \
  --risk-profile aggressive \
  --max-budget-hkd 120000 \
  --financing-preference margin \
  --format json

Generate one decision card:

bash

cd <skill_dir>
uv run --directory runtime/hkipo-next hkipo-next decision-card 2476 --format markdown

Batch explicit symbols:

bash

cd <skill_dir>
uv run --directory runtime/hkipo-next hkipo-next batch 2476 2613 2590 --format json

Batch the current watchlist:

bash

cd <skill_dir>
uv run --directory runtime/hkipo-next hkipo-next batch --watchlist --format json

Output Cues#

• score returns factor contributions, cost breakdown, assumptions, and risk disclosure.
• decision-card adds a headline, position suggestion, and exit plan.
• partial outputs are often still usable; they usually mean source conflicts or incomplete upstream data, not total failure.

Companion Skills#

• Use $hkipo-parameter-manager if there is no active parameter version or the user wants tuning.
• Use $hkipo-review-optimizer after outcomes are known and the user wants to refine the model.