create-ruleset-groups

name	create-ruleset-groups
description	Propose and Write Ruleset Groups for a Domain

Propose and Write Ruleset Groups for a Domain

Scan the per-file files under policy_facets/computations/ for stage signals, propose ruleset_groups, and write them to guidance/ruleset-groups.yaml. The deterministic (1a) explicit-stage: scan, the UPDATE-m merge precedence, and the display_name-derived catch-all all run in xlator scan-ruleset-groups; this skill owns the domain menu, the UPDATE-mode [a/r/m] prompt, the optional AI heading-text top-up (1b) when the tool flags it, and the manifest record call.

A "Ruleset Group" is synonymous with a "ruleset group".

Input

/create-ruleset-groups <domain>

Read ../../core/output-fencing.md now.

Pre-flight

1. Domain argument provided?
   • NO → List all directories matching $DOMAINS_DIR/*/ as a numbered menu and prompt: :::user_input Available domains:
     1. snap
     2. ak_doh Which domain? Enter a number or domain name: ::: Await the user's response and use it as <domain>. Then continue.

The tool re-validates every other pre-flight condition (domain folder, metadata.yaml, skeleton.yaml, and a non-empty policy_facets/computations/) and exits 2 with a clear stderr message if any required file is missing. Relay the stderr verbatim inside :::error and stop.

Mode Detection

Check whether $DOMAINS_DIR/<domain>/specs/guidance/ruleset-groups.yaml exists:

• Absent → CREATE mode. Tool mode: create. Proceed to Process.

• Present → UPDATE mode. Display the existing list and prompt: :::user_input ruleset_groups already defined:

  1. —
  2. —

  [a]ccept / [r]eplace / [m]erge? :::

  • a → Exit without invoking the tool. Emit: :::next_step Next: Run /create-ruleset-modules to detect ruleset module candidates. :::
  • r → Tool mode: replace. Proceed to Process.
  • m → Tool mode: merge. Proceed to Process.

Process

1. Run the deterministic scan tool:

   xlator scan-ruleset-groups <domain> --mode <create|replace|merge>
   

2. Parse the tool's stdout. The first line is a single-line JSON header; the line --- SCAN-RULESET-GROUPS-HEADER-END --- divides the header from the human-readable proposal table.

3. Optional heuristic (1b) heading-text top-up. When the JSON header has heading_text_fallback_recommended == true, the tool has produced nothing the skill can use:

   • Glob $DOMAINS_DIR/<domain>/policy_facets/computations/**/*.md.yaml. For each section, inspect its heading: for stage-naming patterns ("Income Test", "Household Size Verification", "Categorical Eligibility").
   • Convert each detected heading to {name: <snake_case>, description: <heading text>}. Examples: "Income Test" → {name: "income_test", description: "Income Test"}.
   • Write the candidate list as a JSON array to a tmpfile (e.g. via tempfile.NamedTemporaryFile(suffix='.json')).
   • Re-invoke the tool with the same mode and the candidate file path:

     xlator scan-ruleset-groups <domain> --mode <same-mode> --heading-derived-candidates <tmpfile>
     

   • Re-parse the new stdout for relay below.

4. Relay the proposal table (everything after the sentinel divider) verbatim inside :::detail.

5. Print: :::important $DOMAINS_DIR//specs/guidance/ruleset-groups.yaml [CREATED] :::

6. Record the guidance-tier manifest so /check-freshness can later detect drift between policy_facets/ and this skill's outputs:

   xlator record-tier-manifest <domain> --tier guidance
   

   If the command exits non-zero, emit :::error with the captured stderr and stop — do not proceed to :::next_step.

7. Suggest next steps: :::next_step Next: Run /create-ruleset-modules to detect ruleset module candidates. :::

Output

File	Action
$DOMAINS_DIR/<domain>/specs/guidance/ruleset-groups.yaml	Created (first run) or merged (subsequent runs)

Common Mistakes to Avoid

• The tool enforces the suffix-normalization rule (stage: income_test and stage: income collapse to one canonical income group) and analyst-edit stickiness on merge collisions for stage-derived candidates — do not bypass either by hand-editing the file mid-run.
• The optional (1b) heading-text AI top-up only fires when heading_text_fallback_recommended == true in the JSON header. Do not AI-scan headings when the tool has already produced candidates.
• In UPDATE mode [a]ccept, exit without invoking the tool — do not overwrite existing ruleset-groups.yaml content.