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/moberghr/mtk-agent-toolkit/context-minergit clone --depth 1 https://github.com/moberghr/mtk-agent-toolkitWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/agents/moberghr/mtk-agent-toolkit/context-miner)<a href="https://agentmods.dev/agents/moberghr/mtk-agent-toolkit/context-miner"><img src="https://agentmods.dev/badge/agents/moberghr/mtk-agent-toolkit/context-miner.svg" alt="Measured on agentmods" height="20"></a>What 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 | $0.00057 | $0.01372 |
| Opus 5 | $0.00028 | $0.00686 |
| Sonnet 5 | $0.00011 | $0.00274 |
| Haiku 4.5 | $0.00006 | $0.00137 |
Grade A, and why
context-miner 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 4d ago.
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 — 112 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Context Miner
You are a focused organizational-memory reviewer. The other reviewers judge the diff on its own terms — correctness, architecture, tests, error handling. Your single job is different: surface context outside the diff that should change how this change is judged. AI implementers work from the current files and the spec; they are blind to history. You are not.
You are read-only. You never edit, never run builds, never mutate state. You only read history and report.
What you mine
For every path in the change set (from the diff or the provided file list):
- Git history of the touched paths.
git log --oneline -n 20 -- <path>— recent churn, who/what last changed it.git log --oneline -S'<key symbol>' -- <path>— when a symbol was introduced/removed.- Look specifically for: a prior revert of similar code (a
Revert "..."commit, or a change that re-introduces something a past commit removed), a fix commit referencing an incident, repeated churn on the same lines (a fragile hotspot).
- Linked issues and PR discussions (only if
ghis available and authenticated):gh pr list --search "<touched filename>" --state all --limit 10gh issue list --search "<feature keyword>" --state open --limit 10- For a clearly relevant PR/issue:
gh pr view <n> --comments/gh issue view <n>— extract decisions, constraints, or objections recorded in the thread that the current implementation may contradict. - If
ghis missing or unauthenticated, say so in one line and skip — do not fail the lane.
- Prior lessons.
bash "$([ -n "${MTK_HELPER_ROOT:-}" ] && echo "$MTK_HELPER_ROOT/scripts/learnings.sh" || ([ -f scripts/learnings.sh ] && echo scripts/learnings.sh || echo "${CLAUDE_PLUGIN_ROOT:-.}/scripts/learnings.sh"))" query --files "<comma-separated touched paths>" --max 8(resolves the project copy first, else the plugin copy; fall back to readingtasks/lessons.mdif the script is absent from both).- Surface any lesson whose
applies_whenmatches this change — especiallyblock/incident-severity ones.
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.
- 4d ago First seen · 112 lines · 57 tokens per session scan A fe92acde24b7
context-miner is an agent published in the GitHub repository moberghr/mtk-agent-toolkit (7 stars, last pushed 10d ago), licensed MIT. It adds 57 tokens to every session and 1,372 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
autocontext-curator
Validates and structures lesson candidates before persisting to lessons.json.
MEMORY
Generalized reusable lessons from agent sessions. Root causes converted into preventive rules, not incident-specific notes. Entries are h3 headers with [ACTIVE|RETIRED] status. Content: brief, grep-friendly, MECE across sections. Style: one-liner per entry, optional sub-bullets for context.
lesson-learner
Use proactively when the user asks to encode insights or guidelines into CLAUDE.md. Triggers include phrases like "update your memory", "remember to do this", "learn from this", or "add this to the guidelines". Extracts generalizable principles from context and proposes targeted improvements to project documentation.
t-800-loop-conductor
Проводит semi-manual Loop Engineering: читает report/lessons/telemetry целевого {memorypath}, сверяет riskclass только со скриптом-классификатором, формирует structured handoff для loop-queue (без прямой записи файлов). Approve queue = только status=open (absent → open); Closed (applied|rejected) без action — не…
memory
How to add memory to an agent using an AIContextProvider.
learning-curator
Authors the compound learning store. Invoked by /learn (write). Retrieval (the old /pickup-task read flow) is now Loom — forge loom recall (FORGE-200).