Load minimal Project Memory units from .agents/pm files for the current task.
npx skills add https://github.com/AMDResearch/dh_comms --skill pm-load
Please help me install this Agent Skill. Skill: pm-load Source: https://github.com/AMDResearch/dh_comms/tree/main/.agents/skills/pm-load Command: npx skills add https://github.com/AMDResearch/dh_comms --skill pm-load
Load the smallest set of Project Memory units relevant to the current task. This avoids polluting context with unrelated knowledge and keeps token budgets tight.
Identify the task scope and load 1–3 relevant PM units using this skill.
High-intent search queries and answers about installing and using this skill.
This skill identifies the minimal project memory units needed by reading .agents/pm/pm-current-state.md and pm-index.md, selecting 1–3 subsystem units based on task scope. It loads only the focused context required, keeping token budgets tight.
The recommended read order starts with .agents/pm/pm-current-state.md to determine active risks and areas in flux, followed by .agents/pm/pm-index.md to obtain the full unit listing with facets and load-when guidance for selection.
You select the right subsystem units by identifying your task scope and referencing the pm-index.md listing. The skill selects 1–3 subsystem units based on that scope and loads them into working memory for focused context.
You need context loading when your token budgets are tight and you want to avoid polluting context with unrelated knowledge. It enables focused context by loading only the 1–3 relevant subsystem units required for the task.
Yes, project memory context loading works without dependencies or external components. It operates internally by reading the .agents/pm/ directory files to identify and load the smallest relevant PM units into working memory.
AMDResearchConvert reflection report recommendations into approved PM unit structural changes.
fratilanicoLoad and isolate information using a 3-stage hierarchy for long AI sessions.
AMDResearchInitialize Project Memory scaffolding by surveying the codebase and generating starter PM units.
arjunrattan-lab-aiManage design review stages, synthesize feedback, and monitor sign-off statuses.
