<< All versions

Skill v1.0.0

currentAutomated scan100/100
sdsrss/agentsmd/agentsmd-rules
──Details
PublishedNovember 4, 2026 at 02:40 AM
Content Hashsha256:9151fe0acb77f219...
Git SHAb600d663feb8
──Files
Files (1 file, 1.5 KB)
SKILL.md1.5 KBactive
SKILL.md · 32 lines · 1.5 KB

name: agentsmd-rules description: Review agentsmd rule promotion/demotion signals (规则升降级治理) using manifest and eligible telemetry. Use to govern the always-on spec. Not for raw hit listings or automatic spec edits; read-only. metadata: version: "1.0.0"


agentsmd-rules

Use the selected SKILL.md absolute path from the live skills list; never infer it from the process cwd. Define and call the adjacent launcher in the same shell. The launcher preserves selected bundle → manifest-owned standalone → versioned CLI identity checks and exports plugin context only for a verified selected bundle.

bash
SKILL_MD="<selected SKILL.md absolute path from the live skills list>"
agentsmd_skill_run() {
node "$(dirname "$SKILL_MD")/scripts/agentsmd-run.js" "$SKILL_MD" "$@"
}

Cross-reference spec/hard-rules.json with rule-specific eligible/evaluated telemetry. A rule without enough evaluated opportunities is not a demotion candidate.

Run:

bash
agentsmd_skill_run --days=30

Use --project=SUBSTR for the informational local-hits lens; governance verdicts remain cross-project. --include-test explicitly folds test-tagged telemetry into both global and scoped calculations.

Surface opportunity/evaluation counts, demote candidates, and self-enforced rules. Per OPERATOR.md §O2, demotion needs sufficient rule-specific evaluated sessions with zero enforcement; promotion needs repeated real repros. The operator decides; never edit the spec automatically.

All versions