You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Add unified skills/ directory following the agentskills.io specification
for multi-agent support (Claude Code, Cursor, Antigravity, etc.).
Skills included:
- aidd-discover: Product discovery, user journeys, personas
- aidd-task: Epic/task planning with requirements
- aidd-execute: TDD implementation workflow
- aidd-review: Code review with OWASP security checks
- aidd-log: Activity logging for completed work
- aidd-commit: Conventional commit formatting
- aidd-user-test: Human and AI test script generation
Each skill follows the SKILL.md format with:
- aiddCommands frontmatter for command extraction
- Full SudoLang type definitions and functions
- Commands blocks for all workflow commands
- Cross-references to other rules and skills
Added command extraction infrastructure:
- lib/skills-extractor.js: Parse skills and generate command files
- Pre-commit hook: Auto-generate ai/commands/*.md and skills/index.md
- ai/commands/help.md: References skills/index.md for command discovery
Updated AGENTS.md with setup instructions for:
- Claude Code (copy AGENTS.md to CLAUDE.md, symlink skills)
- Cursor (symlink skills to .cursor/skills)
This approach allows a single skills/ directory to work with all
compatible agents via symlinks, maintaining user control over
.cursor and .claude directories.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Copy file name to clipboardExpand all lines: AGENTS.md
+51Lines changed: 51 additions & 0 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -41,3 +41,54 @@ If any conflicts are detected between a requested task and the vision document,
41
41
3. Ask the user to clarify how to resolve the conflict before proceeding
42
42
43
43
Never proceed with a task that contradicts the vision without explicit user approval.
44
+
45
+
## Agent Skills
46
+
47
+
AIDD includes reusable skills in the `skills/` directory following the [agentskills.io](https://agentskills.io) specification. These skills work with any compatible agent (Claude Code, Cursor, etc.).
48
+
49
+
### Available Skills
50
+
51
+
| Skill | Purpose |
52
+
|-------|---------|
53
+
|`aidd-discover`| Product discovery, user journeys, personas |
54
+
|`aidd-task`| Plan and break down epics into tasks |
55
+
|`aidd-execute`| Implement using TDD |
56
+
|`aidd-review`| Code review with security focus |
57
+
|`aidd-log`| Document changes to activity-log.md |
58
+
|`aidd-commit`| Conventional commit formatting |
59
+
|`aidd-user-test`| Generate human and AI test scripts |
60
+
61
+
### Setup for Claude Code
62
+
63
+
Claude Code doesn't natively support `AGENTS.md`. To enable AIDD:
64
+
65
+
```bash
66
+
# Copy or symlink AGENTS.md to CLAUDE.md
67
+
cp AGENTS.md CLAUDE.md
68
+
# or: ln -s AGENTS.md CLAUDE.md
69
+
70
+
# Symlink skills directory
71
+
mkdir -p .claude
72
+
ln -s ../skills .claude/skills
73
+
```
74
+
75
+
### Setup for Cursor
76
+
77
+
```bash
78
+
# Symlink skills directory (Cursor also discovers .claude/skills)
79
+
mkdir -p .cursor
80
+
ln -s ../skills .cursor/skills
81
+
```
82
+
83
+
### Manual Invocation
84
+
85
+
Primary commands invoke their corresponding skill:
86
+
-`/task` - Plan a new epic
87
+
-`/execute` - Implement with TDD
88
+
-`/review` - Code review
89
+
-`/discover` - Product discovery
90
+
-`/commit` - Conventional commits
91
+
-`/log` - Activity logging
92
+
-`/user-test` - Generate test scripts
93
+
94
+
For the complete command reference including sub-commands, see `skills/index.md`.
List commands from please.mdc and report them to the user.
3
+
List available commands and their usage.
4
+
5
+
Read [skills/index.md](../../skills/index.md) for complete command reference and build help text from it.
4
6
5
7
Constraints {
6
-
Before beginning, read and respect the constraints in please.mdc.
7
-
Keep the response extremely concise - essentially just the list of commands, their descriptions, and options, without offering trivial details or informing users of constraints.
8
+
Before beginning, read and respect the constraints in @ai/rules/please.mdc.
9
+
Keep the response extremely concise - essentially just the list of commands, their descriptions, and options, without offering trivial details or informing users of constraints.
0 commit comments