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 skills add mindbox-cloud/agent-skills --skill filter-explaingit clone --depth 1 https://github.com/mindbox-cloud/agent-skillsWrote 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/skills/mindbox-cloud/agent-skills/filter-explain)<a href="https://agentmods.dev/skills/mindbox-cloud/agent-skills/filter-explain"><img src="https://agentmods.dev/badge/skills/mindbox-cloud/agent-skills/filter-explain/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/mindbox-cloud/agent-skills/filter-explain"><img src="https://agentmods.dev/badge/skills/mindbox-cloud/agent-skills/filter-explain.svg" alt="Reviewed on agentmods" width="80" 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 5.1 | $0.00145 | $0.01522 |
| Opus 5.5 | $0.00058 | $0.00609 |
| Sonnet 5.5 | $0.00029 | $0.00304 |
| Haiku 4.5 | $0.00015 | $0.00152 |
Grade A, and why
filter-explain 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 12d 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 — 123 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Filter explain
Choose the project and input
Find the connected server's filter_* tools. Explain a missing connection. Keep calls on
that server: names and IDs identify records only within the project it serves.
Use the complete original filter JSON. A filtered-list URL cannot be imported, and these
tools do not fetch saved filters by name. If only a link or an unseen filter is supplied,
ask for its JSON. For a successful build already in this conversation, reuse its confirmed
finalSql and selected records instead of converting it again.
Freshly read the connected server's maintained, up-to-date entry point: filter_wiki_read
with paths = ["README.md"].
Use its navigation and current tool guidance. Reuse pages within this explanation. Examples below name a tool and the arguments that carry the request; fill in the rest from the schema the server exposes. Use the actual input and returned paths. Tool names may have a server prefix.
1. Convert the original JSON to SQL
Pass the complete JSON unchanged to filter_json_to_sql as its filterJson object — for
example this filter, customers aged 30 and over:
{
"entityType": "User",
"filterFactory": "and",
"innerConditions": [{
"entityType": "User",
"filterFactory": "age",
"value": {"mode": "concrete", "value": {"range": {"from": "30", "to": ""}, "unit": "Years"}}
}]
}
Read the returned sql, selected records and warnings. The tool already attempts to resolve
catalogue IDs to names; do not repeat successful lookups. status: read means the JSON was
read, not that the platform has accepted the filter. unlabelled means some values remain
unnamed; unsupported or UNSUPPORTED_FILTER(...) marks conditions the conversion cannot
explain. Keep the original JSON and these gaps visible in your reasoning. Do not reconstruct
a simpler filter to make them disappear.
2. Read the meaning and resolve remaining labels
Identify the root, fields, relations, operators and business concepts in the SQL. Search the wiki for each distinct concept whose meaning or limitations you need to establish, combining related terms in a regex and reusing relevant pages already read:
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.
- 12d ago First seen · 123 lines · 145 tokens per session scan A ff672ad0ee2d
filter-explain is a skill published in the GitHub repository mindbox-cloud/agent-skills (6 stars, last pushed 9d ago), licensed MIT. It adds 145 tokens to every session and 1,522 once invoked, about $0.0006 per session on Opus 5.5. 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 2026-10-01.
Other skills, from other repositories
prove-the-gates
Audit the repository's own checks — for each test, gate and CI script, establish what it guards, whether anything still runs it, and whether it actually fails when you break what it protects. Use when the user says "prove the gates", "are our tests real", "audit the test suite", "which checks are fake", "תוכיח את…
implementation-notes
Maintain an IMPLEMENTATIONNOTES.md decision log during any multi-step implementation session. Use this in EVERY session that involves writing or modifying code across more than one file, or any session expected to run longer than a few tool calls. Records decisions, deviations from plan, and open questions so nothing…
full-audit
Audit a whole repository in one pass — dispatch every applicable audit lane in parallel, prove the repo's own gates fail when they should, and return one evidence-backed report ranked by severity. Use when the user says "full audit", "audit this repo", "check the whole codebase", "תעשה אודיט מלא", "אודיט לריפו"…
fable-method
The routing logic of the Fable Method - decides which unknowns-discovery technique (blindspot-pass, interview-me, prototype-first, implementation-notes, merge-quiz) applies to the current task, and which to skip. Consult this at the START of any non-trivial task, or when the user says "fable", "the method", "how…
blindspot-pass
Run a blindspot scan to surface the user's unknown unknowns BEFORE implementing a task. Use this whenever the user starts a non-trivial task in a domain they say (or appear) to be unfamiliar with, whenever they say "blindspot", "unknowns", "what am I missing", "before we start", or when a request is vague enough that…
convergence-guard
Stop-loss discipline for implementation sessions - detect fix-loops and runaway token burn, and stop to replan instead of patching. Consult during any debugging or implementation session, especially when tests keep failing, the same file keeps being edited, or the user mentions stoploss, token budget, fix loops, or…