Extending PiSAFE
High-performance AI coding agent CLI written in Rust with zero unsafe code
Overview
From the repository's own README, as read at the audited commit. Badges and raw HTML are left out.
Guide for extending Pi — decide between skills, extensions, prompt templates, themes, context files, or custom models, then create and package them.
Includes nested sub-skills:
- skill-creator — detailed guidance for creating Pi skills
Installation
pi install git:github.com/tmustier/pi-extensions
Or symlink for local development:
ln -s /path/to/pi-extensions/extending-pi ~/.pi/agent/skills/extending-pi
46a2537504c9OBSERVED · 2026-10-10What it tells the agent
The instruction file, verbatim from the audited commit — this is the text the model reads, and the surface the audit's instruction layer examines. Quoted here so you can judge it without cloning anything.
--- name: extending-pi description: Guide for extending Pi — decide between skills, extensions, prompt templates, themes, context files, or custom models, then create and package them. Use when someone wants to extend Pi, add capabilities, create a skill, build an extension, or make a Pi package. --- # Extending Pi Help the user decide what to build and where to find guidance. ## What to build | Goal | Build a... | Where | |------|----------|-------| | Teach Pi a workflow or how to use a tool/API/CLI | **Skill** | Read `skill-creator/SKILL.md` for detailed guidance | | Give Pi a new tool, command, or runtime behavior | **Extension** | Read Pi docs: `docs/extensions.md` | | Reuse a prompt pattern with variables | **Prompt template** | Read Pi docs: `docs/prompt-templates.md` | | Set project-wide coding guidelines | **Context file** | `AGENTS.md` in project root or `.pi/agent/` — just markdown | | Change Pi's appearance | **Theme** | Read Pi docs: `docs/themes.md` | | Add a model or provider | **Custom model** | Read Pi docs: `docs/models.md` (JSON) or `docs/custom-provider.md` (extension) | | Share any of the above | **Package** | Read Pi docs: `docs/packages.md` | ## Skill vs Extension — the fuzzy boundary If `bash` + instructions can do it, prefer a **skill** (simpler, no code to maintain). If you need event hooks, typed tools, UI components, or policy enforcement, use an **extension**. Examples: - "Pi should know our deploy process" → **Skill** (workflow instructions) - "Pi should confirm before `rm -rf`" → **Extension** (event interception) - "Pi should use Brave Search" → **Skill** (instructions + CLI scripts) - "Pi should have a structured `db_query` tool" → **Extension** (registerTool)
Trust audit
SAFEgrade B · trust 89/100 Nothing in the source contradicts what it says it does. Grade A is reserved for packages that have also passed the behavioural sandbox.
| Layer | What it checks | Result |
|---|---|---|
| L0 | Provenance & inventory | WARN |
| L1 | Static analysis of the code | PASS |
| L2 | Instruction surface (what it tells the agent) | PASS |
| L3 | Class-specific surface | PASS |
| L4 | Behavioural (sandbox) | SKIPPED |
What the source does
- Filesystem
- none-observed
- Network
- none-observed
- Shell
- none-observed
- Dependencies
- pinned
- Secrets in source
- none-found
Findings (1)
.beads/beads.jsonl
Gates applied: no_behavioural_pass.
46a2537504c9full audit observations/trust-audit/skill/dicklesworthstone__extending-pi.json · Report an issue / request a re-scanAudit history
Every audit this skill has had.
| Date | Source | Verdict | Grade | Score | Change |
|---|---|---|---|---|---|
| 2026-10-10 | 46a2537504c9 | SAFE | B | 89 | first audit |
Questions
What does the Extending Pi skill do?
High-performance AI coding agent CLI written in Rust with zero unsafe code
Is Extending Pi safe to install?
The audit found nothing in the source that contradicts what it says it does, and graded it B (89/100). Grade A is held back for packages that have also passed a sandboxed behavioural run, which is why a clean skill reads B.
What can Extending Pi access on my machine?
The audit observed no filesystem, network or shell use at all in its source.
How current is this page?
The grade is for one exact copy of the source (46a2537504c9), read on 2026-10-10. The repository is watched, and a new audit runs when it changes — this is the first audit.