skills/profile/SKILL.md
Build, run, and analyze Unity profiler data with perf-report-style call-stack attribution
npx skillsauth add api-haus/my-claude-workflow profileInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
Profile the game with call-stack-level GC allocation and CPU time attribution.
This global skill entry covers methodology cross-references only. The full per-project
implementation lives in each project's own .claude/skills/profile/. Read the project-local
SKILL.md for script inventory, argument grammar, benchmark sections, and pitfall catalogue.
/mnt/archive4/UNITY/Projects/woweyreey/.claude/skills/profile/SKILL.md
Uses unity-cli exec (IPC into live editor) for build. deploy-deck.sh is pure bash./mnt/archive4/UNITY/Projects/zori_test_bed/.claude/skills/profile/SKILL.md
Uses batchmode exclusively — no live editor, no unity-cli. Build step is build.sh <project>.For any profile dispatch targeting zority_6_0 | zority_6_3 | zority_6_4, the build step is:
${CLAUDE_SKILL_DIR}/build-zority.sh zority_6_4
# or absolute path:
${CLAUDE_SKILL_DIR}/build-zority.sh /mnt/archive4/UNITY/Projects/zori_test_bed/zority_6_4
build-zority.sh is the permanent canonical form of the ad-hoc script that was written to
/tmp/build-zority-6-4.sh during the fog-detail-perf Step 9 dispatch (2026-05-18). It is
parameterised: accepts any zority_6_* shorthand or an absolute path.
Do NOT write ad-hoc build scripts to /tmp/ for this testbed. The entry point exists here.
The zori_test_bed CLAUDE.md explicitly removed unity-cli from its skillset:
"The unity-cli (IPC-into-running-editor) tooling is gone from this project's skillset. It was unaffordable: it required keeping an editor process open and focused, and a single domain reload during agent work killed the assumption it relied on."
Every Unity invocation in the testbed goes through:
unity <project> -batchmode -quit [-executeMethod ...] -logFile -
Never pass -nographics to any batchmode invocation in this testbed. From CLAUDE.md:
"Never pass -nographics. Caps RenderTexture at 4096 and breaks anything that allocates GPU resources during import / EditMode fixtures (RenderTexture.Create failed). Applies to compile checks, tests, and -executeMethod runs alike."
The ad-hoc script /tmp/build-zority-6-4.sh correctly omitted -nographics. The testbed
skill's build.sh had this flag present as a latent bug — that's why the profile agent wrote
the ad-hoc script instead of using the existing file. build-zority.sh (this skill) is the
corrected, permanent version.
build-zority.sh applies ulimit -n 8192 before spawning Unity. This works around a Mono
NamedPipeServerStream FD-assertion in Unity 6000.4.x (corefx bugfix), which asserts when an
accepted-socket FD exceeds the cached _SC_OPEN_MAX. 8192 is sufficient for asset import and
IL post-process; it does not restrict the build output.
/build-run (at /mnt/archive4/UNITY/Projects/zori_test_bed/.claude/skills/build-run/SKILL.md)
covers release builds via Zori.TestBed.Editor.BuildScript.BuildLinuxRelease.
build-zority.sh here covers development builds (profiler-enabled) via
Zori.TestBed.Editor.ProfileBuild.BuildDevelopmentLinux. Use /build-run when you want a
production binary; use build-zority.sh (via /profile) when you need a profiler capture.
Both share the same -nographics prohibition and ulimit -n 8192 workaround.
data-ai
Extract research content from YouTube presentations, PDFs, or PPTX files into structured markdown. Dispatches each pass to a dedicated sub-agent (research-extractor / research-vision / research-refiner) so per-deck vision passes scale to hundreds of slides without bloating the parent context.
documentation
Write a handoff prompt for a future session. A handoff is a continuation-link — minimal context plus a kickoff line the user can copy-paste. Never a diagnosis, never an investigation script, never a prescribed deliverable.
testing
Multi-agent orchestration mode. The orchestrator never reads, edits, runs, or tests directly — it scopes work, runs a re-implementation audit, presents a freeform method brief with grounded recommendations, then dispatches every step to sub-agents through shared context files at `docs/orchestrate/<topic>/`. Use when invoked via /delegate, when the user asks to orchestrate or coordinate multi-agent work, or when the task explicitly calls for delegation.
development
Create or switch to a git worktree for isolated feature/fix development