Agent skill

OpenLogi macOS Permissions Triage

by AprilNEA in AprilNEA/OpenLogi

Decides whether an OpenLogi device problem on macOS is a privacy-permission (TCC) problem, using agent log lines, and says which identity needs which grant.

Apache-2.0Auto-check: notesDevelopment

Install OpenLogi macOS Permissions Triage

skills CLI
$ npx skills add AprilNEA/OpenLogi --skill openlogi-macos-permissions -a claude-code

Project install by default; add -g for ~/.claude/skills/.

GitHub CLI
$ gh skill install AprilNEA/OpenLogi openlogi-macos-permissions --agent claude-code

Project scope by default; add --scope user for a personal install. Needs GitHub CLI 2.90.0 or later (public preview).

Manual copy
$ git clone --depth 1 https://github.com/AprilNEA/OpenLogi.git skills-src && mkdir -p .claude/skills && cp -r skills-src/.claude/skills/openlogi-macos-permissions .claude/skills/openlogi-macos-permissions && rm -rf skills-src

Use ~/.claude/skills/ instead of .claude/skills for a personal install. The folder must contain SKILL.md.

Claude Code skills documentation · loads skills from .claude/skills/

Facts

Skill name
openlogi-macos-permissions
GitHub stars
23k
Token cost
~2.5k tokens
SKILL.md length
1,307 words
Files
2 (incl. scripts)
Skills in repo
5
Repo updated
First seen
Licence
Apache-2.0

At a glance

Decides whether an OpenLogi device problem on macOS is a privacy-permission (TCC) problem, using agent log lines, and says which identity needs which grant.

  • Works in 8 steps: First decide whether it is TCC at all → The identity map → Diagnose → …
  • No devices appear in the GUI while `openlogi list` works
  • SKILL.md covers 1. First decide whether it is…, 2. The identity map, 3. Diagnose and 4. Responsible process, plus 4 more sections
  • Runs Shell scripts from its folder

What it does

The guiding idea is that TCC authorizes identities, not processes, so a permission report is a claim rather than a root cause. Before anything else the agent classifies the failure by one question: did the HID++ channel open? A log line with zero candidate interfaces is an enumeration problem, an unsupported or disconnected device, and not TCC. A message that Input Monitoring is not granted is TCC, and the grant goes to the identity the skill names. A message that it is granted but the device cannot open points to another app holding the device, usually Logi Options+, or a stale permission session, and a successful open followed by timeouts is a probe or transport problem.

Most of these log lines are debug-level, so a log captured without `OPENLOGI_LOG=debug` cannot rule anything out, and with no log at all the first job is to get one. The agent also reads the stderr provenance of `openlogi list` to learn whether the result came from the running agent or from the CLI's own code-signing identity and HID stack before treating it as evidence. A bundled `scripts/diagnose.sh` helps collect information, and the skill is meant to be read before changing OpenLogi's permission code, the agent's launch path, the GUI's permission rows or the macOS bundling and signing code.

When your agent uses it

  • No devices appear in the GUI while `openlogi list` works
  • Triaging a Failed to open device or Pairing failed report on macOS
  • Working out which permission OpenLogi needs or why it never prompted
  • Before changing OpenLogi's permission, launch or signing code

Example prompts

  • “The OpenLogi GUI is empty but openlogi list shows my mouse. Is this a TCC problem?”
  • “Triage this agent log: Failed to open device, Input Monitoring is granted to this process.”
  • “Why was I never prompted for Input Monitoring after installing OpenLogi?”

Requirements

  • macOS, where OpenLogi's permission checks apply
  • The OpenLogi agent log, ideally captured with `OPENLOGI_LOG=debug`

Workflow steps

8 steps, taken from the step headings in SKILL.md.

  1. First decide whether it is TCC at all
  2. The identity map
  3. Diagnose
  4. Responsible process
  5. What the APIs actually do
  6. Invariants — do not break these
  7. Check current evidence
  8. What this cannot fix

What it can do on your machine

Read from SKILL.md and the folder at commit ae91d50. It shows what the files ask for, not the result of running them.

  • Tool permissions

    Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.

    From allowed-tools in the SKILL.md frontmatter.

  • Runs code

    Ships 1 file in scripts/ (Shell), which the agent can run.

    From the folder's file list and the shell code blocks in SKILL.md.

  • Network

    No URLs in SKILL.md.

    From URLs in SKILL.md, links to its own repository left out.

  • Credentials

    Names no API keys, tokens, secrets or passwords.

    From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.

Context cost

OpenLogi macOS Permissions Triage loads about 2.5k tokens when it runs. Until then it costs about 184 tokens; SKILL.md has 1,307 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it
~184
When it runs · the whole SKILL.md, loaded when a task matches
~2.5k

Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.

Safety

Auto-check: notes

The automated check noted patterns worth knowing about, such as sudo or a known installer.

  • NoteRuns commands with sudoSKILL.md:129
    sudo launchctl procinfo $(pgrep -x openlogi-agent) | grep -i responsible

Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); the scripts in this folder are not scanned.

SKILL.md

The full file from AprilNEA/OpenLogi at commit ae91d50, republished under its Apache-2.0 licence (© AprilNEA). 1,307 words, ~2,544 tokens.

Download SKILL.mdSave it as .claude/skills/openlogi-macos-permissions/SKILL.md (or your agent's skills folder). This skill also uses 1 other file; get the full folder from GitHub.
name
openlogi-macos-permissions
description
Decide whether a macOS problem is a privacy-permission (TCC) problem, and act on it correctly. Use when triaging a report that no devices appear / the GUI is empty while `openlogi list` works / "Failed to open device" / "Pairing failed" on macOS; when a user asks which permission OpenLogi needs or why they were never prompted; when reasoning about Input Monitoring, Accessibility, kTCCServiceListenEvent, kTCCServiceAccessibility, code-signing identity, responsible process, or bundle identifiers; and before changing anything under openlogi-permissions, openlogi-hid/permissions.rs, the agent's launch/self-restart path, the GUI's settings permission rows, or the macOS bundling and signing code in xtask.

macOS permissions (TCC) in OpenLogi

One sentence to keep: TCC does not authorize processes, it authorizes identities. A permission report is a claim, not a root cause. Identify the failing layer and the identity that performed the operation before diagnosing it.

1. First decide whether it is TCC at all

Three unrelated failures reach the user as the same symptom (an empty device list). Classify before doing anything else. The discriminator is one log line: did the channel open?

Agent logLayerTCC?
HID++ candidate interfaces count=0enumerationNo. The device's HID++ collection never matched — unsupported device, or not connected.
failed to open HID++ channel … Failed to open device: Input Monitoring is NOT granted to this process…openYes. The message classifies itself — grant Input Monitoring to the identity named in §2.
… Failed to open device: Input Monitoring is granted to this process — another app may hold the device exclusively, or macOS is serving a stale permission session (log out and back in)openThe grant is fine. Quit the other app (usually Logi Options+), or log out and back in. See §5.
opened HID++ channel … then Device::new failed / enumerate_features failed with Channel(Timeout) or report writer callback error: 0xE00002D6probeNo. The open succeeded, so investigate the probe/transport path, including the current status of upstream sidit77/async-hid#45.

All of these lines are debug-level except the open failure, which is a warn. A log captured without OPENLOGI_LOG=debug can only ever show the middle row — in such a log, the absence of the other lines is not evidence of anything.

Read openlogi list's stderr provenance before using its result:

  • (inventory read from the running agent) means the HID++ rows came from the same agent snapshot the GUI consumes.
  • (no agent reachable — reading hardware directly; …) or the protocol-version note means the CLI used its own code-signing identity and HID stack because no usable compatible agent snapshot was available.

Camera enumeration is direct in both cases. Without the provenance, a device in openlogi list is not evidence for or against the agent's TCC grant.

If there is no log at all, go to §3 — getting a log is the whole job.

2. The identity map

The long-running app's HID and input grants belong to the agent. Direct CLI operations use the CLI's identity instead.

IdentityBinaryNeeds
org.openlogi.agent…/Contents/Library/LoginItems/OpenLogi Agent.appInput Monitoring (opens HID) and Accessibility (owns the event tap)
org.openlogi.openlogi…/Contents/MacOS/openlogi-desktopCamera only. It is a pure IPC client and needs neither of the above.
openlogi…/Contents/MacOS/openlogi (embedded CLI)Input Monitoring only when list falls back to direct HID or a hardware diagnostic accesses HID directly
org.openlogi.overlay…/LoginItems/OpenLogiOverlay.appNothing

Rules that drive most reports:

  • The identity the user must grant is OpenLogi Agent, which lives inside the app bundle. Bundles built before the rename spell that directory OpenLogiAgent.app; both are the same identity (org.openlogi.agent), and the grant survived the rename because the path is not part of the designated requirement. The System Settings + picker will not browse into a bundle, so they have to use Go-to-Folder. Say this explicitly; do not tell someone to "grant OpenLogi permission".
  • A grant to the GUI does nothing for the agent, and vice versa. There is no bundle-wide grant.
  • Every copy is its own identity. A dev build (org.openlogi.agent-dev), a second install, or a bundle still sitting in ~/Downloads each get their own row. Confirm which binary is actually running before trusting any grant — the diagnose script warns when the running agent is not the one being inspected.

3. Diagnose

Run scripts/diagnose.sh from this skill, or the same steps by hand. It is read-only and safe to hand to a reporter.

Ask for the agent's log file first. launchd discards the agent's stderr, so the agent also writes a daily-rotated file (7 kept) a reporter can attach:

sh
case ${XDG_STATE_HOME:-} in
  /*) state_home=$XDG_STATE_HOME ;;
  *) state_home=$HOME/.local/state ;;
esac
ls "$state_home/openlogi/"

It carries panics too. Only when that file is missing or predates the failure is a foreground run worth its cost:

sh
OPENLOGI_LOG=debug \
  "/Applications/OpenLogi.app/Contents/Library/LoginItems/OpenLogi Agent.app/Contents/MacOS/openlogi-agent"

Note what that costs: run from a terminal, the agent's responsible process becomes the terminal (§4), so the run you are observing is not the run that failed. Compare identities before concluding anything from it — in particular, a successful open under the terminal's grant does not clear the copy launchd runs.

Reading the TCC database directly is not an option for a normal user — it is itself TCC-protected and returns authorization denied without Full Disk Access.

Show full SKILL.md (581 more words)Show less

4. Responsible process

macOS attributes a TCC request to the responsible process, which for a plain child process is the parent. An agent spawned directly by the GUI therefore asks with the GUI's identity, and the user's grant to OpenLogi Agent appears to do nothing.

The launch paths that break the chain live in openlogi-desktop/src/services/ipc/launch.rs and run in this order:

  • registered login item → launchctl kickstart gui/<uid>/<service label> (launchd spawns it directly: its own responsible process, plus crash respawn from the service plist's KeepAlive). The registration itself is SMAppService in openlogi-desktop/src/platform/registration/macos.rs, driven by the launch_at_login setting.
  • packaged helper, not registered → /usr/bin/open -g -n <bundle> (LaunchServices parents it under launchd, so it is its own responsible process — but unsupervised)
  • bare dev binary → disclaim::Command (wraps responsibility_spawnattrs_setdisclaim)

Never spawn a helper with plain std::process::Command on macOS. Check with:

sh
sudo launchctl procinfo $(pgrep -x openlogi-agent) | grep -i responsible

It must name the agent itself. OpenLogi.app or Terminal.app there means every grant on that machine is being ignored.

5. What the APIs actually do

  • IOHIDCheckAccess — queries only. Never prompts, and never registers the app in System Settings, so an app that only ever calls this cannot be granted: the user has no row to tick.
  • IOHIDRequestAccess — prompts. Blocks the calling thread until the user answers, so it must not run on the async runtime. It is also not real-time: after a grant or revoke the calling process keeps seeing the old answer until it restarts. That is why the agent calls binary_watch::relaunch_after_input_monitoring_grant().
  • IOHIDDeviceOpen — denial is silent. There is no TCC-specific error, so the transport pairs every open failure with openlogi_hid::permissions::has_access() and says which case it is (§1). Keep it that way: a bare Failed to open device is not reportable.

6. Invariants — do not break these

  1. Only the agent holds the long-running app's HID/input permissions. Any UI that reports permission state must read it from the agent over IPC, never by querying its own process. Direct CLI diagnostics are a separate, explicit identity and must not be used to infer the agent's grant.
  2. Every helper launch establishes its own responsible process (§4), and the spawn result is checked — a silently failed disclaim leaves the agent running under the GUI's identity, which is invisible until a user reports it.
  3. TCC matches the full designated requirement. A bundle-identifier or signing-identity change can invalidate every existing grant. The Verify production bundle identities step in .github/workflows/build.yml is load-bearing; do not weaken it.
  4. Sign inside-out. The helper needs its own stable designated requirement so its grant survives updates; --deep cannot give it one. See xtask/src/commands/macos/bundle/signing.rs.
  5. Prompt from the process that needs the permission, not from whichever process happens to be running. A TCC grant is scoped to the identity that asked.

7. Check current evidence

Do not carry a "currently fixed/broken" snapshot in this skill. Query the current tracker and latest release notes, identify the reporter's release, and read the matching report, comments, current code, and logs. Treat the report's diagnosis as a claim until the provenance and direct evidence in §§1–4 establish the root cause. Re-check linked upstream transport issues before attributing a probe timeout to them.

8. What this cannot fix

Say so plainly rather than promising a fix:

  • Apple's + picker not browsing into bundles.
  • IOHIDDeviceOpen's silent denial — we can only infer it by checking access separately.
  • A stale tccd decision that needs a full logout, or an MDM policy.
  • Ad-hoc-signed local builds: their designated requirement is cdhash-based, so the grant goes stale on every rebuild. Use an Apple Development identity for dev bundles.

© AprilNEA, Apache-2.0. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file

Files

SKILL.md and 1 other file (scripts) in .claude/skills/openlogi-macos-permissions of AprilNEA/OpenLogi.

  • SKILL.md
  • scripts/diagnose.sh

Open the folder on GitHubat commit ae91d50

Compare with similar skills

OpenLogi macOS Permissions Triage next to the 5 skills that share the most tags, products or categories with it. Stars are the repository's; “used in” counts other GitHub owners with a copy.

OpenLogi macOS Permissions Triage compared with similar skills
SkillStarsUsed inTokensAuto-checkLicenceRepo updated
OpenLogi macOS Permissions Triage this skillAprilNEA/OpenLogi23k—~2.5kAutomated safety check: NotesApache-2.0
Debuggnomeria/usbtree691—~715Automated safety check: PassMIT
Debug Like Expertglittercowboy/taches-cc-resources2k—~2.8kAutomated safety check: PassMIT
Bug Finder for daisyUIsaadeghi/daisyui43k—~2.3kAutomated safety check: PassMIT
Root Cause Debugginggarrytan/gstack136k—~1.4kAutomated safety check: PassMIT
Cmux Debugging Guidemanaflow-ai/cmux28k1 repos~1.1kAutomated safety check: PassCustom licence

Similar skills

  • Debug

    gnomeria/usbtree

    Systematic root-cause debugging — reproduce, isolate, fix at the source, prove the fix.

    691 GitHub stars~715 tokensUpdated 1 mo ago
    DevelopmentAuto-check passed
  • Debug Like Expert

    glittercowboy/taches-cc-resources

    Deep analysis debugging mode for complex issues. An agent skill from glittercowboy/taches-cc-resources.

    2k GitHub stars~2.8k tokensUpdated 6 mo ago
    DevelopmentAuto-check passed
  • Bug Finder for daisyUI

    saadeghi/daisyui

    Investigates suspected bugs in the daisyUI monorepo through read-only analysis, then writes a decision-ready fix plan in tmp/bugs without changing any product code.

    43k GitHub stars~2.3k tokensUpdated 2 days ago
    DevelopmentAuto-check passed
  • Root Cause Debugging

    garrytan/gstack

    Investigates bugs, errors and stack traces in phases and requires a root-cause hypothesis to be confirmed before any fix is written.

    136k GitHub stars~1.4k tokensUpdated yesterday
    DevelopmentAuto-check passed
  • Cmux Debugging Guide

    manaflow-ai/cmux

    Covers debug logging, the Debug menu, profiling rules and runtime pitfalls for working on the cmux macOS terminal app.

    28k GitHub starsUsed in 1 repo~1.1k tokens
    DevelopmentAuto-check passed
  • Codewhale Dogfood Install

    codewhale-hq/Codewhale

    Proves a Codewhale change in the real product: a stamped release build, an atomic local install, fresh-shell verification and manual QA that automated gates cannot cover.

    41k GitHub stars~1.3k tokensUpdated today
    DevelopmentAuto-check passed

More from AprilNEA/OpenLogi

  • Guides recording, privacy review and offline verification of OpenLogi device fixtures with the fixture contribute and verify commands, without treating replay as proof of hardware behavior.

    23k GitHub stars~1.2k tokensUpdated yesterday
    Auto-check passed
  • Testing OpenLogi UI

    AprilNEA/OpenLogi

    Verifies OpenLogi's native GPUI interface with focused tests, the component gallery and a mock agent, choosing the evidence that fits each change.

    23k GitHub stars~1.1k tokensUpdated yesterday
    Auto-check passed
  • Plans the smallest check that could disprove a code change in the OpenLogi project, then escalates through reproduction, focused tests and a final gate before a push.

    23k GitHub stars~1.4k tokensUpdated yesterday
    Auto-check passed
  • OpenLogi Device Diagnosis

    AprilNEA/OpenLogi

    Finds the first failing layer when an OpenLogi Logitech device is missing or misbehaving across enumeration, open, probe, IPC and UI.

    23k GitHub stars~1.6k tokensUpdated yesterday
    Auto-check passed

Works with

Categories

Questions about OpenLogi macOS Permissions Triage

What does OpenLogi macOS Permissions Triage do?

Decides whether an OpenLogi device problem on macOS is a privacy-permission (TCC) problem, using agent log lines, and says which identity needs which grant. The guiding idea is that TCC authorizes identities, not processes, so a permission report is a claim rather than a root cause. Before anything else the agent classifies the failure by one question: did the HID++ channel open?

When should I use OpenLogi macOS Permissions Triage?

OpenLogi macOS Permissions Triage fits situations like: no devices appear in the GUI while `openlogi list` works; triaging a Failed to open device or Pairing failed report on macOS; working out which permission OpenLogi needs or why it never prompted; before changing OpenLogi's permission, launch or signing code.

How do I install OpenLogi macOS Permissions Triage in Claude Code?

Run `npx skills add AprilNEA/OpenLogi --skill openlogi-macos-permissions -a claude-code`. Or copy the skill folder (.claude/skills/openlogi-macos-permissions in AprilNEA/OpenLogi) into .claude/skills/openlogi-macos-permissions in your project. Claude Code loads it when a task matches its description.

How do I install OpenLogi macOS Permissions Triage in Codex?

Run `npx skills add AprilNEA/OpenLogi --skill openlogi-macos-permissions -a codex`. Or copy the skill folder (.claude/skills/openlogi-macos-permissions in AprilNEA/OpenLogi) into .agents/skills/openlogi-macos-permissions in your project. Codex loads it when a task matches its description.

Can I use OpenLogi macOS Permissions Triage in Cursor, Gemini CLI or GitHub Copilot?

Cursor, Gemini CLI, GitHub Copilot and OpenCode also load SKILL.md folders. With the skills CLI, run `npx skills add AprilNEA/OpenLogi --skill openlogi-macos-permissions -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/openlogi-macos-permissions, .gemini/skills/openlogi-macos-permissions, .github/skills/openlogi-macos-permissions and .opencode/skills/openlogi-macos-permissions in your project.

What does OpenLogi macOS Permissions Triage need to run?

Going by SKILL.md and its folder, OpenLogi macOS Permissions Triage needs a shell for the scripts in its folder. Our summary lists: macOS, where OpenLogi's permission checks apply; The OpenLogi agent log, ideally captured with `OPENLOGI_LOG=debug`.

Does OpenLogi macOS Permissions Triage access the network?

SKILL.md contains no URLs. Any network use would come from the scripts or tools the agent runs. This is read from the text; nothing was executed.

Is OpenLogi macOS Permissions Triage safe to install?

Our automated static check of SKILL.md found notes only (runs commands with sudo), nothing it rates as a warning. It is not a guarantee. The check reads SKILL.md only: the scripts in the folder are not scanned, so read them before running anything.

What licence does OpenLogi macOS Permissions Triage use?

OpenLogi macOS Permissions Triage is published under the Apache-2.0 licence (the repository's licence). It allows redistribution, so the full SKILL.md is shown on this page.

How many tokens does OpenLogi macOS Permissions Triage use?

About 2.5k tokens (SKILL.md is roughly 10k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full.

What are the alternatives to OpenLogi macOS Permissions Triage?

Skills that share tags, products or a category with OpenLogi macOS Permissions Triage: Debug (gnomeria/usbtree, 691 stars), Debug Like Expert (glittercowboy/taches-cc-resources, 2k stars), Bug Finder for daisyUI (saadeghi/daisyui, 43k stars) and Root Cause Debugging (garrytan/gstack, 136k stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.

Who maintains OpenLogi macOS Permissions Triage?

AprilNEA (a GitHub user) maintains it in AprilNEA/OpenLogi, which has 23,278 GitHub stars. The repository holds 5 skills in this directory. The repository was last updated on October 10, 2026.

Source: AprilNEA/OpenLogi on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.