Getting it into your agent
It runs from inside its repository, so the clone comes first — what it calls does not travel with the file alone.
git clone --depth 1 https://github.com/leopu00/job-hunter-teamnpx agentmods add skills/leopu00/job-hunter-team/cache-pruneWrote 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/leopu00/job-hunter-team/cache-prune)<a href="https://agentmods.dev/skills/leopu00/job-hunter-team/cache-prune"><img src="https://agentmods.dev/badge/skills/leopu00/job-hunter-team/cache-prune/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/leopu00/job-hunter-team/cache-prune"><img src="https://agentmods.dev/badge/skills/leopu00/job-hunter-team/cache-prune.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector warn
SkillSpector: 1 finding, up to high
These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →
- high YARA Match · line 3 YARA rule matched a hack tool or exploit indicator (offensive tools, reconnaissance, privilege escalation, or exploit frameworks).Fix: Remove offensive tool references and exploit code. Legitimate agent skills should not contain penetration testing tools, exploit frameworks, or reconnaissance utilities.
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 | $0.00103 | $0.01286 |
| Opus 5.5 | $0.00041 | $0.00514 |
| Sonnet 5.5 | $0.00021 | $0.00257 |
| Haiku 4.5 | $0.00010 | $0.00129 |
Grade A, and why
cache-prune 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 1mo 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 — 87 lines — stays where its author put it; the contents beside it link to each section on GitHub.
cache-prune — reclaim shared caches
The shared $JHT_HOME accumulates two caches that grow monotonically until reclaimed:
| Path | What it stores | Typical growth (sample 2026-05-02) |
|---|---|---|
$JHT_HOME/.cache/uv/ |
wheel cache for every uv pip install |
~364 MB |
$JHT_HOME/.codex/logs_2.sqlite |
Codex telemetry SQLite (71% TRACE rows) | ~223 MB |
Neither is needed on disk: uv re-downloads if it has to, Codex truncates TRACE rows safely. The numbers above came from a continuous run; on a fresh $JHT_HOME they start at 0 and reach hundreds of MB within a few days.
The single safe command
node /app/cli/bin/jht.js cache prune
Idempotent and no-op when there's nothing to reclaim. Internally:
uv cache prune— drops stale wheels (keeps the active set referenced by current installs).- SQLite
VACUUMonlogs_2.sqliteafter deleting old TRACE rows. - Cleanup of ephemeral Codex temp files.
Each step has a safety gate: idle > 1h on the operations that are destructive (VACUUM lock, TRACE delete) — if the team is actively burning tokens the step is skipped.
When to run
- 👨⚕️ End of a routine Dottore round (~24h of continuous run, or at the start of an idle operational day).
- 📉 On demand if
du -sh $JHT_HOME/.cache $JHT_HOME/.codexshows growth > 800 MB total. - 🚫 NEVER mid-budget-critical (proj > 95%) — the VACUUM 30s blocks the Codex SQLite the Sentinel reads through the bridge.
- 🚫 NEVER in reaction to a Sentinel
[ORDINE]— orders demand pacing/scaling actions, not housekeeping.
Safety: what NOT to touch
The team has other caches that look similar but are NOT in scope here:
| Path | Why hands-off |
|---|---|
.cache/ms-playwright/ |
browser binaries pinned by version — re-downloading is slow + flaky |
.cache/claude-cli-nodejs/ |
Anthropic CLI runtime cache, recreated lazily but bigger when warm |
$JHT_HOME/logs/ |
Sentinel state lives here. Wiping it loses the EMA window and several minutes of monitoring history. |
What ships with it
6 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- 1mo ago First seen · 87 lines · 103 tokens per session scan A ef8b64eaf5a9
cache-prune is a skill published in the GitHub repository leopu00/job-hunter-team (52 stars, last pushed today), licensed MIT. It adds 103 tokens to every session and 1,286 once invoked, about $0.0004 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 2026-08-30.
Other skills, from other repositories
freehire-job-search
Use when finding jobs for someone — searching or filtering the freehire catalogue by keyword, region, seniority, skills, salary or company; reading a posting or a company's open roles; discovering what values a filter accepts; reading the saved job-search profile before asking someone what they want; handing over a…
freehire-mail-triage
Use when sorting a job seeker's application mail into the freehire tracker from their own mail client (himalaya, mbsync, notmuch, the Gmail API, any IMAP) — pushing a fetched batch, judging what each message is, linking it to an application, advancing a stage from a reply, or draining the two queues the matcher fills…
freehire-tailor-cv
Use when reframing someone's CV toward a specific vacancy through freehire — reading the fit analysis for a tailored copy, fetching or editing the CV document by path, rendering the ATS PDF, or deciding what may and may not be written into a CV. Carries the evidence rule: anything stating what the candidate DID needs…
freehire-market-fit
Use when measuring a set of skills — a CV's stack, or one skill on its own — against live open-vacancy demand on freehire; when telling a candidate which in-demand skills they are missing and what learning each would unlock; or when gauging how much of a filtered market (a role, a region, a seniority) their current…
freehire-track-applications
Use when recording or reviewing what someone applied to on freehire — marking a job applied, bookmarking one for later, moving an application through its stages (applied through offer, or rejected/withdrawn), attaching a note, or listing what they are tracking. Also covers reporting a posting that never answered them…
kent-workflows
How to use the Kent CLI to author and edit Kent workflow graphs — nodes, edges, transitions, parameters, context modes, and project links. Use when the user asks to define, build, inspect, or modify Kent workflows.