Release v1.9.7: move agent skills back to .claude/skills so they are discoverable

.agents/skills/capture-tui-diagram/SKILL.md → .claude/skills/capture-tui-diagram/SKILL.md

@@ -71,14 +71,14 @@ sleep 6                            # let the next poll apply it, THEN capture
 lines, trailing pad whitespace, and scrollbar thumb artifacts:
 
 ```bash
-python .agents/skills/capture-tui-diagram/clean_capture.py tmp/raw.txt > tmp/grid.txt
+python .claude/skills/capture-tui-diagram/clean_capture.py tmp/raw.txt > tmp/grid.txt
 ```
 
 For a **single-column** capture (the `stack` preset) add `--compress` to collapse
 runs of blank chart rows so a ~47-row frame fits a doc:
 
 ```bash
-python .agents/skills/capture-tui-diagram/clean_capture.py tmp/stack_raw.txt --compress > tmp/stack.txt
+python .claude/skills/capture-tui-diagram/clean_capture.py tmp/stack_raw.txt --compress > tmp/stack.txt
 ```
 
 **Never `--compress` a side-by-side frame** (grid, or dashboard+table): the two

.agents/skills/publish-cover/SKILL.md → .claude/skills/publish-cover/SKILL.md

@@ -68,7 +68,7 @@ Gate 1–4  →  Build  →  Audit  →  [MANUAL CONFIRM]  →  Deploy  →  Ver
 ### Step 1 — Build
 
 ```bash
-bash .agents/skills/publish-cover/scripts/build-cover-dist.sh
+bash .claude/skills/publish-cover/scripts/build-cover-dist.sh
 ```
 
 What the script does deterministically:

.agents/skills/record-tui-gif/SKILL.md → .claude/skills/record-tui-gif/SKILL.md

@@ -30,7 +30,7 @@ tape. actop monitors the whole system, so that's fine.
 From anywhere in the repo:
 
 ```bash
-bash .agents/skills/record-tui-gif/record.sh
+bash .claude/skills/record-tui-gif/record.sh
 ```
 
 This drives a GPU workload against the llama.cpp router (llamacpp-first default;
@@ -53,14 +53,14 @@ Examples:
 
 ```bash
 # Default: llama.cpp router, OpenAI protocol, original qwen model:
-bash .agents/skills/record-tui-gif/record.sh
+bash .claude/skills/record-tui-gif/record.sh
 # Heavier load, longer generations:
-CONCURRENCY=3 NUM_PREDICT=8192 bash .agents/skills/record-tui-gif/record.sh
+CONCURRENCY=3 NUM_PREDICT=8192 bash .claude/skills/record-tui-gif/record.sh
 # Ollama fallback (native /api/generate, agentic profile):
 ROUTER_URL=http://localhost:11433 MODEL=qwen3.6:35b-a3b-agentic API=ollama \
-  bash .agents/skills/record-tui-gif/record.sh
+  bash .claude/skills/record-tui-gif/record.sh
 # Record a custom tape with idle gauges (e.g. a quick layout check):
-SKIP_WORKLOAD=1 TAPES=tmp/mytape.tape bash .agents/skills/record-tui-gif/record.sh
+SKIP_WORKLOAD=1 TAPES=tmp/mytape.tape bash .claude/skills/record-tui-gif/record.sh
 ```
 
 ## The GPU workload driver
@@ -71,11 +71,11 @@ generations across N concurrent workers until Ctrl-C or `--duration`.
 
 ```bash
 # llama.cpp (default): OpenAI protocol against the llamacpp router
-.venv/bin/python .agents/skills/record-tui-gif/gpu_workload.py \
+.venv/bin/python .claude/skills/record-tui-gif/gpu_workload.py \
     --url http://localhost:9040 --model qwen3.6:35b-a3b \
     --api openai --concurrency 2 --num-predict 4096
 # ollama fallback: native /api/generate (llama.cpp 404s on this path)
-.venv/bin/python .agents/skills/record-tui-gif/gpu_workload.py \
+.venv/bin/python .claude/skills/record-tui-gif/gpu_workload.py \
     --url http://localhost:11433 --model qwen3.6:35b-a3b-agentic \
     --api ollama --concurrency 2 --num-predict 4096
 ```

.agents/skills/record-tui-gif/actop-demo.tape → .claude/skills/record-tui-gif/actop-demo.tape

@@ -15,7 +15,7 @@
 # the dashboard drops to ~80 cols and silently degrades mid-take; 1800px
 # (~212 cols) leaves it ~122 cols, so grid survives the whole recording.
 #
-# Render from the repo root:  vhs .agents/skills/record-tui-gif/actop-demo.tape
+# Render from the repo root:  vhs .claude/skills/record-tui-gif/actop-demo.tape
 # (or via record.sh, which drives a GPU workload so the gauges move).
 
 Output images/actop-demo.gif

.agents/skills/record-tui-gif/record.sh → .claude/skills/record-tui-gif/record.sh

@@ -11,7 +11,7 @@
 #   5. stop the GPU workload
 #
 # Run from anywhere inside the repo:
-#   bash .agents/skills/record-tui-gif/record.sh
+#   bash .claude/skills/record-tui-gif/record.sh
 #
 # Config via env:
 #   TAPES="tmp/a.tape tmp/b.tape"   tapes to record (default: bundled actop-demo.tape)

CHANGELOG.md

@@ -6,6 +6,22 @@ This project follows a Keep a Changelog-style format and uses version tags for r
 
 ## [Unreleased]
 
+## [1.9.7] - 2026-08-13
+
+### Fixed
+
+- **The agent skills were invisible to the agent.** `record-tui-gif`,
+  `publish-cover`, `capture-tui-diagram`, and `run-actop` were moved from
+  `.claude/skills/` to `.agents/skills/` in the 1.4.x docs reorg. Claude Code
+  discovers project skills only under `.claude/skills/`, so from that point on
+  none of the four were ever offered — they stayed correct, committed, and
+  unused, and got followed only when someone happened to grep them out of the
+  tree. `audit-conformance` stayed behind in `.claude/skills/` and kept working,
+  which is what made the breakage hard to notice.
+  All four are moved back, their internal `.agents/skills/...` path references
+  rewritten, and `CLAUDE.md` now states the constraint so a future reorg does
+  not silently re-break discovery.
+
 ## [1.9.6] - 2026-08-12
 
 ### Fixed

CLAUDE.md

@@ -136,6 +136,8 @@ The `docs/` directory contains essential system reviews, research, and operation
 - `docs/SPEC-system.md`: Detailed system design reference — native bindings, sampling layer, SoC profile fallback, TUI rendering, testing contract. Kept in sync with the code on every PR that touches architecture.
 - `docs/ROADMAP.md`: Open hardware/metric-coverage gaps and their priority.
 
+**Agent skills live in `.claude/skills/` — nowhere else.** Claude Code discovers project skills only under `.claude/skills/` (plus `~/.claude/skills/` and plugins); a skill parked anywhere else is invisible to the agent and will simply never be invoked. `record-tui-gif`, `publish-cover`, `capture-tui-diagram`, and `run-actop` were moved to `.agents/skills/` in a past docs reorg and silently stopped being offered — they were still correct, still committed, and still useless, because nothing surfaced them. Do not "consolidate" them elsewhere; if another tool needs its own convention, give that tool a copy or a symlink and leave `.claude/skills/` as the real home.
+
 **Conformance auditing:** the `/audit-conformance` skill (`.claude/skills/audit-conformance/`) periodically judgment-scans the whole tree against 12 coding rules (layering, dead code, DRY, naming, swallowed errors) — the whole-codebase counterpart to diff-scoped `/code-review`. It writes an actionable `docs/TODO-conformance-YYYY-MM-DD.md` and never proposes structural/guard tests (they would violate the functional-tests-only mandate).
 
 ## Coding Style & Naming Conventions

docs/cover.html

@@ -165,7 +165,7 @@
 
 <section class="hero">
   <div class="wrap">
-    <span class="badge">v1.9.6</span>
+    <span class="badge">v1.9.7</span>
     <h1>actop<span>.</span></h1>
     <p class="tagline">
       Watch your Apple Silicon Mac the way it actually works — <strong>sudoless</strong>,
@@ -274,7 +274,7 @@ <h2>actop — Apple <span style="color: var(--accent);">C</span>hip top</h2>
 
 <footer>
   <div class="wrap">
-    <p>actop v1.9.6 · an independent project inspired by <a href="https://github.com/tlkh/asitop">asitop</a> · MIT licensed</p>
+    <p>actop v1.9.7 · an independent project inspired by <a href="https://github.com/tlkh/asitop">asitop</a> · MIT licensed</p>
   </div>
 </footer>
 

pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "actop"
-version = "1.9.6"
+version = "1.9.7"
 description = "Apple Silicon (M1–M4) power, GPU, ANE & memory-bandwidth monitor — sudoless TUI + Python API for profiling local LLM / MLX / CoreML inference"
 readme = "README.md"
 license = { text = "MIT" }