writing-plans + subagent-driven-development Setup¶
Source: nousresearch/hermes-agent Skills.sh: writing-plans | subagent-driven-development
1. Installation¶
npx skills add nousresearch/hermes-agent --skill writing-plans
npx skills add nousresearch/hermes-agent --skill subagent-driven-development
2. Prerequisites¶
| Requirement | Details |
|---|---|
| Hermes Agent | v0.16.0+ with delegate_task tool enabled |
| Subagent config | delegation.max_concurrent_children set in config.yaml |
| No external API keys | Both skills are local-only |
3. Capabilities¶
writing-plans¶
| Capability | Trigger | Output |
|---|---|---|
| Task decomposition | "write a plan for X" | Numbered task list with file paths |
| Dependency mapping | Complex multi-file features | Dependency graph with execution order |
| Code scaffolding | "scaffold the plan" | Stub files with signatures |
| Effort estimation | Plan review | Per-task complexity scores |
subagent-driven-development¶
| Capability | Trigger | Output |
|---|---|---|
| Parallel execution | "execute the plan" | Spawned subagents per task |
| 2-stage review | After each subagent completes | Review → approve/reject → retry |
| Result reconciliation | After all subagents complete | Merged results, conflict resolution |
| Rollback on failure | Subagent failure | Clean revert of failed task artifacts |
4. CLI / Command Reference¶
# Load skills in a session
# The agent auto-loads these when relevant -- no manual trigger needed
# Manual invocation pattern:
# 1. Ask Hermes to write a plan
# "Write an implementation plan for adding dark mode to the dashboard"
# → writing-plans skill activates, produces numbered task list
# 2. Execute the plan
# "Execute the dark mode plan"
# → subagent-driven-development activates, spawns workers
# 3. Review progress
# "Show me the status of the dark mode implementation"
# → Agent reconciles subagent results, shows completion state
5. CorpusIQ Use Cases¶
| Use Case | How It Works |
|---|---|
| Feature development | Break new CorpusIQ features into plans, execute in parallel -- 3-5x faster than sequential |
| Docs updates | Plan doc page creation across multiple sections, execute simultaneously |
| System audit fixes | Decompose audit findings into fix tasks, delegate to subagents |
| Connector integration | Plan connector setup (MCP server + docs + tests), execute in parallel |
| Multi-platform content | Plan social posts across platforms, delegate per-platform formatting |
6. Troubleshooting¶
| Issue | Solution |
|---|---|
| Subagent context missing files | Ensure context field in delegate_task includes all relevant file paths |
| Subagent results conflict | Review the reconciliation output -- agent flags conflicts for manual resolution |
| Plan too granular | Ask "simplify the plan to 3-5 tasks" -- plans auto-scale to feature complexity |
| Subagent timeout | Increase delegation.timeout in config.yaml; split oversized tasks |
Verification:
# Verify skills installed
npx skills list | grep -E "writing-plans|subagent-driven-development"
# Test plan generation
# In a Hermes session: "Write a plan for adding a hello-world endpoint"
# Should produce: numbered tasks with file paths and dependencies
Part of the Hermes Skills Library -- 133+ agent skills. Built by CorpusIQ.
Part of the Hermes Skills Library -- 133+ agent skills. Built by CorpusIQ.¶
*
This Hermes repo is one of the largest structured collections of public AI, automation, business, and technology documentation. Content remains attributed to original authors and repositories. Indexed and organized by www.CorpusIQ.io.