Inspecting managed skills and automation output
SkillAI & modelsUse when listing or reading agent-managed automation skills, viewing automation run artifacts, or inspecting skills owned by the standard Hermes user install without mutating them.
Available today. Use it from your connected AI after setup.
No other account needed.
Connect ahel once, and every AI you use reads what you have installed.
Then ask your AI: use the Inspecting managed skills and automation output skill
What this skill tells your AI
The instructions your AI receives, as published by scriptedalchemy/tracedecay in plugin/skills/inspecting-managed-skills/SKILL.md and read by ahel’s review.
The daemon automation loop (skill writer, memory curator, session reflector) drafts managed skills and records durable run artifacts. This skill is the read-only window into that state; every lifecycle change (approve, disable, archive, install) goes through the tracedecay automation CLI or the dashboard instead.
Workflow
- List managed skills →
tracedecay_skill_list(state?: filter by lifecycle state,include_body?): metadata, lifecycle state, usage summary, and stale/archive/improvement evidence for every agent-managed skill in the active profile. Start here to see what automation has produced. - Read one skill →
tracedecay_skill_view(idrequired,include_support_files?): full metadata, body markdown, usage summary, and support files for a single managed skill. Use before recommending approval, edits, or archival. - Read a run artifact →
tracedecay_automation_run_artifact_view(run_id,kind: e.g.traces,feedback,generated_evals,validation_gate,optimizer_diagnosis,codex_handoff): the hash-verified JSON payload of one durable automation run artifact. Find run ids viatracedecay automation runs listwhen needed. - Inspect standard Hermes skills →
tracedecay_hermes_skill_bridge(include_skill_bodies?,include_pending_payloads?): read skills, pending approval records, usage telemetry, and archive counts from the one supported~/.hermesinstall. The tool accepts no profile or home selector.
Guardrails
- All four tools are read-only; none of them approve, edit, or delete anything. For lifecycle changes hand the user the matching CLI commands:
tracedecay automation skills approve|disable|archive|restore <id>andtracedecay automation skills install --target <host> --output <path>. Hermes-owned lifecycle changes stay in Hermes. - Managed skills are distinct from this bundled skill set: they live in the TraceDecay profile store, not in the plugin. Do not edit bundled plugin skills based on managed-skill evidence.
Handoff
- Running or configuring the automation jobs themselves →
tracedecay automation run/tracedecay automation config(CLI). - Inspecting session-reflection fact automation outcomes →
tracedecay automation facts list|view|apply|reject(CLI; mutate only on explicit request). - Memory fact curation →
tracedecay:project-memory.
If tools are deferred or MCP fails
- Deferred (names listed without schemas): load once with ToolSearch —
select:tracedecay_skill_list,tracedecay_skill_view,tracedecay_automation_run_artifact_view,tracedecay_hermes_skill_bridge— then call normally. - MCP error/timeout/disconnect: same tool, same args, via shell:
tracedecay tool skill_list(seetracedecay:using-the-cli). Never query.tracedecaydatabases directly; never abandon the graph over transport.
Deliverable
Do not end this workflow without: the requested skill list, skill body,
artifact payload, or standard Hermes inventory, plus the exact CLI command for any
lifecycle action the user should take next. Report any tracedecay_metrics:
line to the user.
Signals
- GitHub stars
- 73
- Forks
- 5
- Last commit
- Sep 2026
Advanced
- Catalog kind
- skill
- Gateway key
inspecting-managed-skills- Source
- github.com/scriptedalchemy/tracedecay