tes-cortex

TES Cortex

/tes-cortex is the preferred shared TES trigger for Cortex memory operations. /tes:cortex is a compatible TES intent alias if the host reports it as an invalid slash. The agent remains the executor; scripts and MCP tools are oracles.

Intent Map

User intent	Preferred action
recall/query/search memory	use read-only cortex_recall MCP or cortex.py recall
read a cell	use read-only cortex_read_cell MCP or cortex.py read-cell
check memory health	run cortex.py verify and cortex.py audit
rebuild recall	run cortex.py rebuild; SQLite remains derived
curate memory quality	run read-only cortex_curate_plan MCP or cortex.py curate-plan --backend lexical
learn from a source	run cortex.py learn or absorb-plan; propose only
close a work cycle	run cortex.py reflect; propose only
promote a durable claim	run cortex.py apply only with explicit approval and evidence

Rules

• Treat sources/**, cells/**, MAP.md, TRAIL.md, LINKS.md, and CONTRACT.md as the memory.
• Treat .tes/cortex/recall.sqlite as rebuildable cache.
• Treat .tes/cortex/semantic.sqlite as rebuildable curation cache.
• Prefer MCP for read-only recall/read/curate/reflect when available.
• Never write sources/** after import.
• Never promote loose summaries; cells need ## Claim and ## Evidence.
• When reflect.curation_due=true, run curate-plan before proposing memory compaction, split, merge, or rejection.
• Never run apply --yes without explicit user authorization.