95 lines
2.7 KiB
Markdown
95 lines
2.7 KiB
Markdown
# Agent Skills for Agentic Engineering
|
|
|
|
20 production-ready SKILL.md files following the Agent Skills open standard. Compatible with Claude Code, Pi Agent, OpenCode, Hermes, and OpenClaw.
|
|
|
|
## Quick Install
|
|
|
|
```bash
|
|
# Via npx (standard)
|
|
npx skills add https://github.com/fdsa/agentic-skills --skill security-foundation
|
|
|
|
# Via bash
|
|
bash install.sh security
|
|
```
|
|
|
|
## Skills
|
|
|
|
### Security Foundation — 6 skills
|
|
- `l3-blacklist` — Bash blacklist hook
|
|
- `l4-whitelist` — Bash whitelist hook (architectural)
|
|
- `l5-no-bash` — Custom tools only
|
|
- `damage-control` — 3-level access control
|
|
- `agent-sandbox` — E2B sandbox isolation
|
|
- `acip-defense` — Prompt injection defense
|
|
|
|
### Multi-Agent Orchestration — 6 skills
|
|
- `agent-team` — Dispatcher pattern
|
|
- `agent-chain` — Pipeline pattern
|
|
- `mental-model` — Agent-owned expertise files
|
|
- `domain-lock` — Per-agent permissions
|
|
- `tilldone` — Task discipline
|
|
- `p2p-coms` — Peer-to-peer communication
|
|
|
|
### Verifier Pro — 5 skills
|
|
- `verifier-builder` — Two-agent observer
|
|
- `confidence-ladder` — PERFECT to FAILED grading
|
|
- `claim-decomposition` — Atomic verification units
|
|
- `read-only-surface` — Read-only tool enforcement
|
|
- `escalation-handler` — Max 3 correction loops
|
|
|
|
### Task Discipline — 4 skills
|
|
- `tilldone-core` — Task list gating
|
|
- `tilldone-progress` — Live progress widget
|
|
- `tilldone-nudge` — Nudge on incomplete
|
|
- `purpose-gate` — Session intent declaration
|
|
|
|
### Autoresearch — 5 skills
|
|
- `experiment-loop` — Run, measure, log, decide
|
|
- `integrity-guard` — Code hashing, grind detection
|
|
- `median-over-best` — Statistical comparison
|
|
- `reward-hack-defense` — Timing integrity
|
|
- `log-to-jsonl` — Structured experiment logging
|
|
|
|
### Observability — 4 skills
|
|
- `tool-call-tracer` — SQLite tool logging
|
|
- `cost-tracker` — Per-session cost tracking
|
|
- `session-replay` — Session debugging
|
|
- `loop-detector` — Loop detection alerts
|
|
|
|
### CEO Board — 11 agents + verifier + tracker
|
|
- 8 board members + CEO + verifier + tracker
|
|
|
|
## Individual Skills
|
|
|
|
| Skill | Invoke | Description |
|
|
|-------|--------|-------------|
|
|
| Agent Tutor | `/agent-tutor` | Socratic tutor for agentic engineering |
|
|
| Feynman Technique | `/feynman` | Deep learning via simplification |
|
|
| Plan F3 | `/planf3` | HTML-first engineering implementation plans from disler/planf3 |
|
|
|
|
## Format
|
|
|
|
All skills follow the Agent Skills open standard:
|
|
|
|
```yaml
|
|
---
|
|
name: skill-name
|
|
description: What this skill does and when to use it
|
|
user-invocable: true
|
|
---
|
|
|
|
# Skill Title
|
|
|
|
[Instructions for the agent to follow]
|
|
|
|
$ARGUMENTS
|
|
```
|
|
|
|
## Listing
|
|
|
|
Skills are listed at `officialskills.sh` (coming soon).
|
|
|
|
## Repo
|
|
|
|
Source: [github.com/fdsa/agentic-skills](https://github.com/fdsa/agentic-skills)
|