Writes actionable markdown implementation plans to .hermes/plans/ without executing code.
npx skills add https://github.com/xu1713/openhorse --skill plan-xu1713
Please help me install this Agent Skill. Skill: plan Source: https://github.com/xu1713/openhorse/tree/main/openhorse/openhorse/skills/software-development/plan Command: npx skills add https://github.com/xu1713/openhorse --skill plan-xu1713
.hermes/plans/, working across local, docker, ssh, modal, and daytona backends. - Use Case: Before refactoring a module, invoke plan mode to get a saved document listing exact file paths, test targets, and verification steps for later execution. ## Quick Start Ask the agent to plan the implementation of your feature and save the plan under .hermes/plans/ without modifying any code.High-intent search queries and answers about installing and using this skill.
Invoke plan mode and describe the task; the agent inspects the repository with read-only commands and writes a markdown plan to .hermes/plans/. No files are edited and no mutating commands are run during the planning turn.
A good plan includes the goal, current context and assumptions, proposed approach, step-by-step tasks, files likely to change, tests and validation steps, plus risks and open questions. For code tasks it should list exact file paths and test targets.
Plans are saved as timestamped markdown files under .hermes/plans/ relative to the active workspace, using the format YYYY-MM-DD_HHMMSS-slug.md. This path works consistently across local, docker, ssh, modal, and daytona backends.
Plan mode only permits read-only inspection commands and writing the plan markdown file itself. It explicitly blocks code edits, mutating terminal commands, commits, pushes, and external actions.
If the task is clear, the plan is written directly; if no instruction accompanies the command, the task is inferred from conversation context. When genuinely underspecified, the agent asks a brief clarifying question instead of guessing.
danalexilewisSynthesize end-of-chat context and docs into a reviewer-only investigation plan.
alphabeenAudit a codebase against a plan document and produce a PASS/FAIL review.
commontoolsincBreak down plans into subtasks and track progress using beads (bd).
HonigbartWrite executable plans to .ai/plans/<slug>/final_plan.md.