Shannon Exec: Autonomous Task Execution

Execute tasks autonomously with automatic library discovery, 3-tier functional validation, and atomic git commits.

Usage

/shannon:exec "task description"
/shannon:exec "task" --dry-run
/shannon:exec "task" --interactive

What It Does

1. Discovers Libraries: Searches npm, PyPI, etc. for existing solutions
2. Generates Code: Uses /shannon:wave for AI code generation
3. Validates Functionally: Runs build + tests + actual functionality checks
4. Commits Atomically: Only validated code enters git history
5. Retries on Failure: Rolls back and retries (max 3x)

Workflow

@skill exec
  task: {user_task}
  options:
    dry_run: false
    interactive: false
    max_iterations: 3

The exec skill orchestrates:

• Phase 1: Context preparation (/shannon:prime)
• Phase 2: Library discovery (shannon discover-libs)
• Phase 3: Task analysis (optional)
• Phase 4: Execution planning
• Phase 5: Code generation + validation + commit loop
• Phase 6: Execution report

Examples

Simple Task

/shannon:exec "create hello.py that prints hello world"

Result: hello.py created, validated, committed in ~20s

Feature with Libraries

/shannon:exec "add authentication to React app"

Result: Discovers next-auth, uses it, validates, commits in ~5min

Dry-Run Mode

/shannon:exec "build e-commerce platform" --dry-run

Result: Shows plan (libraries, steps) without executing

Requirements

• Shannon CLI V3.5.0+ (provides executor modules)
• Serena MCP (required for caching)
• Clean git state (no uncommitted changes)

Output

• Git branch with validated commits
• Execution report saved to Serena
• Libraries used documented
• Validation results per commit

---

See skills/exec/SKILL.md for complete workflow details.