GPT-5 finding
Enabled-skill counter includes commented lines; contradicts "Never count from comments" and can misclassify POWER
mediumbughigh
- skills/fork-cohort/SKILL.md:89-91
- skills/fork-cohort/SKILL.md:281-283
- aeon.yml:1-5
The implementation uses a plain grep for "enabled: true" without filtering out commented lines, yet the Constraints section asserts that comments are skipped. Grep does not ignore comments by default, so any '# enabled: true' in aeon.yml would be counted, inflating ENABLED_COUNT and potentially upgrading forks to POWER incorrectly. This directly affects cohort classification and notifications.
Recommendation
Filter out commented lines (and whitespace-only lines) before counting, or parse YAML structurally. Examples: - POSIX tools: ENABLED_COUNT=$(awk '/^[[:space:]]*#/ {next} /enabled:[[:space:]]*true/ {c++} END {print c+0}' /tmp/fork-aeon.yml) - Grep-only: ENABLED_COUNT=$(grep -E "^[[:space:]]*[^#].*\benabled:[[:space:]]*true\b" /tmp/fork-aeon.yml | wc -l | tr -d ' ') - Prefer robust YAML parsing if yq is available: ENABLED_COUNT=$(yq -r '.skills | to_entries | map(select(.value.enabled == true)) | length' /tmp/fork-aeon.yml)