Lab: Footprint decision drill

#

Goal

#

For three feature ideas, choose a ladder rung, justify it in ≤5 bullets, and list files you would touch — without implementing them.

Prereqs

#

• Read Footprint Ladder and Skills, plugins, and tools

Steps

#

For each scenario below, write:

1. Chosen rung (1–6)
2. Why lower rungs are insufficient (or why this rung is enough)
3. Files/paths you would touch (user space vs core)
4. One cache/session risk to watch

Scenario A — Daily GitHub digest

#

“Every morning, summarize my open PRs and post to Telegram.”

Scenario B — Custom weather API

#

“My company has an internal weather HTTP API with a private token; the agent should call it with structured args.”

Scenario C — React to Discord messages

#

“When someone reacts with 👀, the bot should acknowledge in-thread.”

Expected outcome

#

A short writeup (markdown note is fine) with three decisions. Suggested reference answers (compare after you write yours):

Your answers may differ slightly; the grading bar is reasoning quality, not identical labels.

Verify

#

1. None of your plans start with “add to _HERMES_CORE_TOOLS” without exhausting lower rungs
2. Scenario B does not put the API token in config.yaml
3. Scenario C does not propose check_fn based solely on HERMES_DESKTOP

Common pitfalls

#

• Treating “agent should do X” as automatic justification for a core tool
• Putting cron logic inside a skill with no durable scheduler
• Solving Discord UX inside run_agent.py

Stretch

#

• Re-do Scenario A with no_agent script-only cron vs full agent cron — when is each better?
• See website/docs/user-guide/features/cron.md and website/docs/guides/

Next lab

#

Cache-safe change review