Define the minimum workflow required to resume work after a fresh /new session without rereading the whole codebase.
Harnessing is currently:
recovery-ready- not yet
fully automatic
That means a new session can recover direction and active work quickly, but it still relies on a deliberate startup routine.
open consumer repo
-> open Harnessing repo
-> ingest latest docs and memory artifacts
-> query for current status and next target
-> generate a lean or work bundle
-> continue implementation
For a consumer such as ontology-for-cm:
- open the consumer repo and Harnessing repo
- run
python src/harnessing/cli.py ingest - run a query such as:
current statusnext engineering targetrevit backend actions
- generate a bundle:
--mode leanfor fast recovery--mode workfor active implementation
- continue work using the recovered context
- avoids rereading large parts of the codebase
- reloads direction, status, and troubleshooting context quickly
- keeps token usage lower than raw doc or transcript injection
- automatic startup bundle generation
- project-specific alias presets
- transcript and tool-write prioritization strong enough for every session
- guaranteed one-command continuity
- add consumer-specific startup presets
- promote more transcript content into reusable memories
- add stronger alias and synonym handling for common project questions
- add export formats that can be injected into Codex and Claude Code with less manual steering