How do I install the Launchd Secret Linter skill?

Run npx skillmds@latest add etanhey/launchd-secret-linter in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Launchd Secret Linter safe to use?

This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Launchd Secret Linter?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Launchd Secret Linter free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Launchd Secret Linter?

EtanHey (@etanhey) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Launchd Secret Linter by etanhey — Agent Skill

Name: Launchd Secret Linter
Author: EtanHey

File contents

launchd-secret Linter

A launchd plist that hardcodes a secret in EnvironmentVariables leaks it into the user-agent environment on first load and contaminates the whole fleet — the gen-18 specimen was a raw GOOGLE_API_KEY leaked exactly this way. A hardened plist carries only an op:// 1Password reference or $VAR env-indirection, never the literal. This skill makes that rule a mechanical, replayable gate.

Scope

The scan target is the plist EnvironmentVariables block.

Run it

python3 lint_cli.py ~/Library/LaunchAgents/*.plist   # exit 1 if any plist hardcodes a secret
python3 -m pytest tests/                              # the RED→GREEN gate (10 cases)

The CLI never prints secret values — only the plist, the env key, and the rule.

Rules (precision-biased — flag literals, never indirection)

Rule	Catches
`raw-secret-value`	A value matching a known provider key shape — Google `AIza…`, OpenAI `sk-…`, GitHub `ghp_/gho_/ghs_/ghu_…`, Slack `xox…`, AWS `AKIA…`, a PEM `-----BEGIN … PRIVATE KEY-----` — that is a literal (not `op://…`, not `$VAR`).
`literal-in-secret-named-var`	An env var whose name says secret (`_API_KEY`, `_TOKEN`, `_SECRET`, `PASSWORD`, `CREDENTIAL`, `PRIVATE_KEY*`) whose value is a non-empty literal that is not an `op://` reference or `$`-indirection.

op:// references and $VAR / ${VAR} indirection are always treated as the hardened (non-leaking) form and never flagged.

The hardened form

<key>GOOGLE_API_KEY</key>
<string>op://Private/google/api_key</string>   <!-- 1Password reference -->
<key>GH_TOKEN</key>
<string>$GH_TOKEN</string>                       <!-- env indirection -->

Do NOT auto-rotate

When this linter finds a hardcoded secret, flag it — do not rotate it automatically. Per gen-18 D7: untrack != redaction; assess the secret's live/billed status before any rotation, and migrate the plist to an op:// reference. Auto-rotation can break live services and is the user's call.

Validated against real artifacts

Run against the live ~/Library/LaunchAgents/ set, the linter both confirms hardened plists clean and surfaces real hardcoded secrets by key (values never printed). The RED→GREEN fixtures (fixtures/) pin the gate; provider-pattern detection is exercised with values built in test code so the committed suite carries no real-shaped secret.

---name: launchd-secret-linterdescription: Lint launchd plists for hardcoded secrets. Triggers: launchd plist, LaunchAgents, plist secret, raw API key in plist, secrets hygiene, op:// reference.---# launchd-secret LinterA launchd plist that hardcodes a secret in `EnvironmentVariables` **leaks it into theuser-agent environment on first load and contaminates the whole fleet** — the gen-18specimen was a raw `GOOGLE_API_KEY` leaked exactly this way. A hardened plist carries onlyan **`op://` 1Password reference** or **`$VAR` env-indirection**, never the literal. Thisskill makes that rule a mechanical, replayable gate.## ScopeThe scan target is the plist `EnvironmentVariables` block.## Run it```bashpython3 lint_cli.py ~/Library/LaunchAgents/*.plist # exit 1 if any plist hardcodes a secretpython3 -m pytest tests/ # the RED→GREEN gate (10 cases)```The CLI never prints secret values — only the plist, the env key, and the rule.## Rules (precision-biased — flag literals, never indirection)| Rule | Catches ||------|---------|| `raw-secret-value` | A value matching a known provider key shape — Google `AIza…`, OpenAI `sk-…`, GitHub `ghp_/gho_/ghs_/ghu_…`, Slack `xox…`, AWS `AKIA…`, a PEM `-----BEGIN … PRIVATE KEY-----` — that is a literal (not `op://…`, not `$VAR`). || `literal-in-secret-named-var` | An env var whose **name** says secret (`*_API_KEY`, `*_TOKEN`, `*_SECRET`, `*PASSWORD*`, `*CREDENTIAL*`, `*PRIVATE_KEY*`) whose value is a non-empty literal that is not an `op://` reference or `$`-indirection. |`op://` references and `$VAR` / `${VAR}` indirection are always treated as the hardened(non-leaking) form and never flagged.## The hardened form```xml<key>GOOGLE_API_KEY</key><string>op://Private/google/api_key</string> <key>GH_TOKEN</key><string>$GH_TOKEN</string> ```## Do NOT auto-rotateWhen this linter finds a hardcoded secret, **flag it — do not rotate it automatically**.Per gen-18 D7: `untrack != redaction`; assess the secret's live/billed status before anyrotation, and migrate the plist to an `op://` reference. Auto-rotation can break liveservices and is the user's call.## Validated against real artifactsRun against the live `~/Library/LaunchAgents/` set, the linter both confirms hardenedplists clean and surfaces real hardcoded secrets by key (values never printed). TheRED→GREEN fixtures (`fixtures/`) pin the gate; provider-pattern detection is exercised withvalues built in test code so the committed suite carries no real-shaped secret.

Frequently asked questions

Launchd Secret Linter