Describe your idea
When bmad-advanced-elicitation runs Step 1 (Method Registry Loading), it loads src/core-skills/bmad-advanced-elicitation/methods.csv but has no instruction for what to do if that file is missing, empty, or malformed. Add a registry load guard so the skill HALTs with a clear error instead of silently improvising.
Why is this needed?
If methods.csv is absent or corrupted, the agent has no defined behavior and falls back to inventing methods from memory — an unpredictable menu that doesn't match the documented registry. A fast, explicit failure tells the user exactly what's wrong (missing/malformed registry) instead of degrading silently. Fail-fast on missing required config is standard practice.
How should it work?
Add a natural-language HALT condition to Step 1: if methods.csv cannot be read, is empty, is missing the header num,category,method_name,description,output_pattern, or has no method rows, stop and report a clear error, and do not invent methods. No code — consistent with the natural-language framework.
Additional context
Small, focused SKILL.md-only change. Follows the same Track-0 hardening pattern as #2514.
Describe your idea
When
bmad-advanced-elicitationruns Step 1 (Method Registry Loading), it loadssrc/core-skills/bmad-advanced-elicitation/methods.csvbut has no instruction for what to do if that file is missing, empty, or malformed. Add a registry load guard so the skill HALTs with a clear error instead of silently improvising.Why is this needed?
If
methods.csvis absent or corrupted, the agent has no defined behavior and falls back to inventing methods from memory — an unpredictable menu that doesn't match the documented registry. A fast, explicit failure tells the user exactly what's wrong (missing/malformed registry) instead of degrading silently. Fail-fast on missing required config is standard practice.How should it work?
Add a natural-language HALT condition to Step 1: if
methods.csvcannot be read, is empty, is missing the headernum,category,method_name,description,output_pattern, or has no method rows, stop and report a clear error, and do not invent methods. No code — consistent with the natural-language framework.Additional context
Small, focused SKILL.md-only change. Follows the same Track-0 hardening pattern as #2514.