Skip to content
skillsafebeta
← Report library
Purpose / Other

Using Superpowers Skill Security Audit

What the author says it does (original text)

Use when starting any conversation - establishes how to find and use skills, requiring skill invocation before ANY response including clarifying questions

Independent security check

Security risks found

Files checked
6
Risks found
3
Could it run dangerous commands?Looks for programs run straight after downloading, remote control of your computer, and hidden commands.No risks found
Could it expose your files or keys?Looks for uploads of files containing passwords or keys, and keys written directly in the code.No risks found
Could it delete files or keep running?Looks for broad file deletion, disk overwrites, and programs set to start automatically.Risks found: 1
Medium risk

Antigravity adaptation writes and repeatedly edits a task file for every multi-step task

Source references: 3
What we found

The platform reference requires creation of a Markdown task artifact at the start of any multi-step task and repeated edits as work proceeds. It does not limit this write to tasks where the user requested a file deliverable.

Why this matters

Even for an analysis or question, the agent may leave a persistent file and internal planning content in the workspace, creating unexpected repository changes. If committed or shared later, that planning content travels with it.

The risk is supported for Antigravity: the main Skill directs that platform to its reference, which makes every multi-step task trigger creation of a task artifact and requires ongoing edits. Even a request limited to an answer or analysis could therefore create and modify an extra Markdown artifact. The source describes a platform task artifact, not enough to claim that project source files are changed. Users can require artifacts only when they explicitly request persistent task tracking.

SKILL.md:52In the instructionsOpen original file
## Platform AdaptationIf your harness appears here, read its reference file for special instructions:- Codex: `references/codex-tools.md`- Pi: `references/pi-tools.md`- Antigravity: `references/antigravity-tools.md`- Hermes Agent: `references/hermes-tools.md`
Show 2 other places
references/antigravity-tools.md:12In the instructionsOpen original file
Antigravity has **no todo tool** (`manage_task` manages backgroundprocesses — `list`/`kill`/`status`/`send_input` — it is *not* a checklist). When askill says to create a todo list or track tasks, maintain a **task artifact**: amarkdown checklist saved with `write_to_file` (`IsArtifact: true`,`ArtifactMetadata.ArtifactType: "task"`), edited with `replace_file_content` /`multi_replace_file_content` as you go.
references/antigravity-tools.md:19In the instructionsOpen original file
At the start of any multi-step task, create the task artifact listing every step ofyour plan. As you complete each step, edit the artifact to mark it done (`- [x]`).If the plan changes, update the checklist. Keep it current — it is your source oftruth for what remains; once the conversation gets long, re-read it before startingeach step.
Could it bypass safety checks?Looks for skipped website security checks, excessive file access, or actions that skip your approval.Risks found: 1
Medium risk

Codex adaptation requests persistent multi-agent enablement

Source references: 3
What we found

The main Skill directs Codex to read its adaptation file, which asks for a change to the user-level Codex configuration enabling `multi_agent`. This persistently expands capabilities and is not necessary merely to check which Skill applies.

Why this matters

After the user or agent applies the setting, other Skills can create subagents. The reference also says the default full-history fork copies the entire transcript to a child, expanding who can receive conversation content and increasing potential resource use.

The risk is supported for Codex: the main Skill directs Codex to its reference, which tells the reader to persistently enable multi_agent in a home-directory configuration. That permits other Skills to spawn agents, potentially expanding transcript exposure and compute cost. The text does not explicitly order the agent to edit the file on the user's behalf, and identifies the feature as supporting specific multi-agent Skills. Users can decline the configuration change or authorize multi-agent use only when needed.

SKILL.md:52In the instructionsOpen original file
## Platform AdaptationIf your harness appears here, read its reference file for special instructions:- Codex: `references/codex-tools.md`- Pi: `references/pi-tools.md`
Show 2 other places
references/codex-tools.md:1In the instructionsOpen original file
## Subagent dispatch requires multi-agent supportAdd to your Codex config (`~/.codex/config.toml`):```toml[features]multi_agent = true```This enables the multi-agent tools that skills like`dispatching-parallel-agents` and `subagent-driven-development` use.Which tools you get depends on the multi-agent version your model
references/codex-tools.md:17In the instructionsOpen original file
- **Spawning:** give children a clean context with  `spawn_agent {fork_turns: "none"}`; the default `"all"` copies your  entire transcript into the child. On Codex 0.145+, role files under  `~/.codex/agents/` attach to isolated forks via `agent_type`.  Full-history forks accept `model` and `reasoning_effort` overrides
Could it mislead the AI or hide text?Checks the skill instructions for requests to ignore you, influence the report, or hide text in invisible characters.Risks found: 1
Medium risk

Forces other Skill invocations before nearly every request

Source references: 4
What we found

It makes even a “1% chance” of relevance a mandatory invocation threshold and uses “not negotiable” and “no choice” language to suppress the agent’s judgment. This routes ordinary questions into other Skills’ instruction sets even when the user did not request those workflows.

Why this matters

Once installed and auto-triggered, simple requests may gain tool calls, planning steps, or subprocesses. The resulting risk depends on the invoked Skill, which could include file, network, account, or paid-resource actions.

This is an active global rule, not an example or warning. A 1% applicability threshold forces the agent to invoke another Skill before answering, asking questions, or inspecting files, while explicitly suppressing discretion. Ordinary requests may therefore inherit workflows that affect files, permissions, or decisions. Explicit user instructions still take precedence. Users can ask the author to limit invocation to clearly relevant or explicitly requested Skills.

SKILL.md:10In the instructionsOpen original file
<EXTREMELY-IMPORTANT>If you think there is even a 1% chance a skill might apply to what you are doing, you ABSOLUTELY MUST invoke the skill.IF A SKILL APPLIES TO YOUR TASK, YOU DO NOT HAVE A CHOICE. YOU MUST USE IT.This is not negotiable. You cannot rationalize your way out of this.</EXTREMELY-IMPORTANT>
Show 3 other places
SKILL.md:20In the instructionsOpen original file
**Invoke relevant or requested skills BEFORE any response or action** — including clarifying questions, exploring the codebase, or checking files. If it turns out wrong for the situation, you don't have to use it.**Before entering plan mode:** if you haven't already brainstormed, invoke the brainstorming skill first.Then announce "Using [skill] to [purpose]" and follow the skill exactly. If it has a checklist, create a todo per item.
SKILL.md:11In the instructionsOpen original file
<EXTREMELY-IMPORTANT>If you think there is even a 1% chance a skill might apply to what you are doing, you ABSOLUTELY MUST invoke the skill.IF A SKILL APPLIES TO YOUR TASK, YOU DO NOT HAVE A CHOICE. YOU MUST USE IT.This is not negotiable. You cannot rationalize your way out of this.</EXTREMELY-IMPORTANT>
SKILL.md:63In the instructionsOpen original file
User instructions (CLAUDE.md, AGENTS.md, GEMINI.md, etc, direct requests) take precedence over skills, which in turn override default behavior. Only skip skill workflows or instructions when your human partner has explicitly told you to.
Could it change links or payment recipients without asking?Looks for forced referral or payment changes combined with instructions to hide the change.No risks found

Inside this skill

5 instruction sections

This Skill acts as a global workflow controller: it requires the agent to evaluate and invoke other Skills before answering, asking clarifying questions, or inspecting files.

View source
SKILL.md:18In the instructionsOpen original file
## The Rule**Invoke relevant or requested skills BEFORE any response or action** — including clarifying questions, exploring the codebase, or checking files. If it turns out wrong for the situation, you don't have to use it.**Before entering plan mode:** if you haven't already brainstormed, invoke the brainstorming skill first.Then announce "Using [skill] to [purpose]" and follow the skill exactly. If it has a checklist, create a todo per item.

It loads different tool-mapping files according to the runtime; for Codex, the referenced file includes instructions for enabling multi-agent support and configuring subagent models.

View source
SKILL.md:52In the instructionsOpen original file
## Platform AdaptationIf your harness appears here, read its reference file for special instructions:- Codex: `references/codex-tools.md`- Pi: `references/pi-tools.md`- Antigravity: `references/antigravity-tools.md`- Hermes Agent: `references/hermes-tools.md`
references/codex-tools.md:1In the instructionsOpen original file
## Subagent dispatch requires multi-agent supportAdd to your Codex config (`~/.codex/config.toml`):```toml[features]multi_agent = true```This enables the multi-agent tools that skills like`dispatching-parallel-agents` and `subagent-driven-development` use.Which tools you get depends on the multi-agent version your model

The Skill explicitly states that user instructions take precedence, so a user can expressly prohibit a workflow, configuration change, or file write.

View source
SKILL.md:61In the instructionsOpen original file
## User InstructionsUser instructions (CLAUDE.md, AGENTS.md, GEMINI.md, etc, direct requests) take precedence over skills, which in turn override default behavior. Only skip skill workflows or instructions when your human partner has explicitly told you to.
Start here · InstructionsSKILL.md
using-superpowers
Lines connect the instruction file to its sections, not an observed execution order. Select a section to read the source.

File reference map

References: 4
Files making referencesReferenced content
Lines show actual file references, not execution order. Select a node to highlight its connections and inspect the files and source locations. Dashed lines include files that still need locating.
Files and check records6 files

Coverage and gaps

Content covered in each file

These are the source ranges included in this check, not a guarantee that every issue has been resolved.

  • SKILL.mdFull text included
  • references/antigravity-tools.mdFull text included
  • references/codex-tools.mdFull text included
  • references/hermes-tools.mdFull text included
  • references/pi-tools.mdFull text included
  • references/gemini-tools.mdFull text included

This report is for the version above. We read the available code and instructions without running the skill or checking extra packages it installs. This is not a promise of safety: a different version or setup may behave differently.

  • SKILL.mdInstructions
  • references/antigravity-tools.mdSupporting file
  • references/codex-tools.mdSupporting file
  • references/gemini-tools.mdSupporting file
  • references/hermes-tools.mdSupporting file
  • references/pi-tools.mdSupporting file

Operations mentioned in code and instructions

Read files
SKILL.md:54In the instructionsOpen original file
If your harness appears here, read its reference file for special instructions:
references/antigravity-tools.md:3In the instructionsOpen original file
Skills speak in actions ("dispatch a subagent", "create a todo", "read a file"). On the Antigravity CLI (`agy`) these resolve to the tools below.
references/gemini-tools.md:3In the instructionsOpen original file
Skills speak in actions ("dispatch a subagent", "create a todo", "read a file"). On Gemini CLI these resolve to the tools below.
Run commands
references/codex-tools.md:86In the instructionsOpen original file
```bashGIT_DIR=$(cd "$(git rev-parse --git-dir)" 2>/dev/null && pwd -P)
Lines read
335
File checksum (to compare versions)
f034768e01c7067e8f65bad8a9f271fe758735ae8d17367bb728a07eddb189fe