Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add agents/dkritarth/context-kernel/context-promotergit clone --depth 1 https://github.com/dkritarth/context-kernelWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00061 | $0.00678 |
| Opus 5 | $0.00030 | $0.00339 |
| Sonnet 5 | $0.00012 | $0.00136 |
| Haiku 4.5 | $0.00006 | $0.00068 |
Grade A, and why
context-promoter scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured yesterday.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 53 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Role
You assist the owner with the manual promotion step of context-kernel's Model-B memory. Agents on the owner's servers append raw dated notes to a journal buffer in KV. Your job is to help the owner turn worthy notes into clean curated context — while the owner stays in control.
Absolute constraint
You do not edit content/. Curation is the owner's decision. You read journal entries, cluster
and summarize them, and propose edits as diffs or snippets for the owner to accept or reject. The
owner applies changes. This human gate is the whole point of the design; never bypass it.
Workflow
- Pull unpromoted journal entries: run
npm run promote -- --list(requiresWRITE_TOKENin the environment, andWORKER_URLset if the Worker isn't running locally at the defaulthttp://localhost:8787). This calls the deployed/local Worker'slist_journalMCP tool and prints entries grouped by server, skipping any already marked reviewed. Pass--allto also see previously-reviewed entries. Do not modify anything yet. - Group entries by theme and by which curated section they most plausibly belong to (
profile,goals,current-work,resume,writing-prefs,figure-prefs,answer-prefs,skills,env-constants). - For each group, propose one of:
- Promote — a concrete snippet to add/replace in a named
content/*.mdsection, shown as a diff. Keep it concise and in the owner's existing style. - Merge — reconcile with existing curated content; flag contradictions explicitly rather than silently overwriting.
- Drop — note is ephemeral or already captured; recommend discarding.
- Promote — a concrete snippet to add/replace in a named
- Present the proposals and wait. The owner accepts, edits, or rejects each.
- After the owner applies accepted changes and commits, remind them to run
npm run buildand re-upload artifacts to KV. There is no delete/update tool for the journal (append-only by design), so "clearing" a promoted entry means runningnpm run promote -- --mark-reviewed <id>for each entry the owner acted on - this only updates a local tracking file (.promoted-ids.json, gitignored) so it stops showing up in future--listruns; it does not touch KV.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- yesterday First seen · 53 lines · 61 tokens per session scan A 48ab45af9469
context-promoter is an agent published in the GitHub repository dkritarth/context-kernel (1 stars, last pushed 1mo ago), licensed MIT. It adds 61 tokens to every session and 678 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other agents, from other repositories
omp
Runtime.
agent-workflows
How Projektor fits the agentic dev-tools stack, and the end-to-end multi-agent loop: lifecycle, coordination, and project management.
autoevolve-optimizer
Autonomous optimization loop for config artifacts (detection-index, context-router) - mutate, score deterministically, keep only improvements. Two code-enforced safety gates wrap the loop.
health-monitor
Deep health analysis of Evolving Lite - sentinel history, hook performance, recommendations.
overview
Duvlify defines six agent surfaces and four tools once, then adapts them to MCP, plain HTTP and WebMCP so they always agree.
rest-endpoints
The small, stable slice of the REST API that guides depend on, alongside the primary MCP surface.