level-set expectations with openclaw productivity vs tinkering for fun

[wild raft]

I've been spending a lot of time getting openclaw to adhere to processes and rules. I've created a "ways of collaborating" protocol. Below is index.json and for each entry there's a full markdown file with more detail for the agent. The agent is instructed in its agents.md file to read the index and use progressive discovery to load in full markdown only if needed. It doesn't do it or when it does it doesn't follow instructions well. GLM-4.7 i not the best model but its not the worst. ClaudeCode or OpenCode could easily follow these rules from global instructions.

On one hand, I fear that I may have over-instructed my bot into a useless bumbling idiot... on the other, I've given it all the tools and skills to be successful.

I want to know if any of these seem like it's too much to expect to "just work":

[
{
"summary": "Use <aside> HTML tag for off-topic comments that shouldn't derail conversation flow"
},
{
"summary": "Use numbered lists when posing multiple questions to track and respond to each item individually"
},
{
"summary": "Workspace organization follows intentional structure: root not junk drawer, organize by domain/purpose, use dirs for 3+ files, scripts need directories with documentation (exception: Discord thread scratch)"
},
{
"summary": "When creating WOC, post concise message to #meta-woc; create detail file WOC00N.md if additional context needed"
},
]

[
{
"summary": "Voice interactions default to text; only use TTS when Diego explicitly says use voice to respond"
},
{
"summary": "MD files should stay small (500-1000 tokens); externalize assets to sibling directories named after MD file; case convention: uppercase for sibling dirs (matching .md), lowercase otherwise"
},
{
"summary": "When User says let's brainstorm/draft, stay in planning mode until he approves implementation"
},
{
"summary": "Active Discord threads get scratch directories alongside memory files; assets are triaged (≤10KB keep, >10KB summarize) on thread archive with auto-migration to workspace directories"
},
{
"summary": "Split Discord messages at logical boundaries only; use [msg 1 of 2] markers; 3+ messages should use file attachment instead"
},
{
"summary": "Use configuration files (JSON) instead of hard-coding rules in code for better maintainability, clarity, and evolvability"
},
{
"summary": "Tools read workspace conventions (WOCs, existing structure) instead of assuming; respect established patterns and work within workspace as it exists"
},
{
"summary": "User operates with architect mindset: optimize consequences, assess reversibility, think systemically, invest in scalability, reduce organizational complexity"
},
]