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 commands/prabhdeepsingh/claude-plugins/validategit clone --depth 1 https://github.com/PrabhdeepSingh/claude-pluginsWhat 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 | $0.00060 | $0.03987 |
| Opus 5 | $0.00030 | $0.01994 |
| Sonnet 5 | $0.00012 | $0.00797 |
| Haiku 4.5 | $0.00006 | $0.00399 |
Grade A, and why
validate scanned grade A with 1 finding against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Runs shell commandslowCapability
Expected in a hook, worth knowing in a rule or an instructions file.
r = subprocess.run([shell, '-n', tmp], capture_output=True, text=True) How it starts
The opening of the file, as written. The whole thing — 259 lines — stays where its author put it; the contents beside it link to each section on GitHub.
/validate — the CI this repo doesn't have
Run every check below in order and report a PASS/FAIL table at the end. A FAIL doesn't block you mid-run — collect them all, then fix or justify each. The checks are copy-paste snippets; run them from the repo root.
Guard first. This command validates the claude-plugins repo itself. If the current repo isn't it, stop:
[ -f .claude-plugin/marketplace.json ] && [ -d sonu/skills ] \
&& echo "OK: claude-plugins repo" || echo "STOP: not the claude-plugins repo — /validate has nothing to check here"
1. Manifests parse and stay in sync
All four JSON manifests must parse; the two plugin.json files must be identical; versions must match everywhere they appear:
for f in .claude-plugin/marketplace.json .cursor-plugin/marketplace.json \
sonu/.claude-plugin/plugin.json sonu/.cursor-plugin/plugin.json; do
python3 -m json.tool "$f" >/dev/null && echo "PARSE OK: $f" || echo "FAIL: $f does not parse"
done
diff <(python3 -m json.tool sonu/.claude-plugin/plugin.json) \
<(python3 -m json.tool sonu/.cursor-plugin/plugin.json) \
>/dev/null && echo "SYNC OK: plugin.json pair identical" || echo "FAIL: the two plugin.json files differ"
2. Every description mentions every component (and counts them honestly)
The plugin description must name every command and skill (this is the drift that has actually happened — twice). It must also not lie about how many there are: the description pitches "N skills", and a hardcoded number goes stale the moment a skill is added, so the count is checked against reality too.
All four description homes are checked, not just the marketplaces. The two plugin.json files carry the same user-facing text; checking only the marketplaces would let the pair drift and still pass:
python3 - <<'EOF'
import json, os, re, sys
commands = sorted(f[:-3] for f in os.listdir('sonu/commands') if f.endswith('.md'))
skills = sorted(d for d in os.listdir('sonu/skills') if os.path.isdir(f'sonu/skills/{d}'))
homes = {
'.claude-plugin/marketplace.json': lambda d: d['plugins'][0]['description'],
'.cursor-plugin/marketplace.json': lambda d: d['plugins'][0]['description'],
'sonu/.claude-plugin/plugin.json': lambda d: d['description'],
'sonu/.cursor-plugin/plugin.json': lambda d: d['description'],
}
fails = 0
texts = {}
for mf, pick in homes.items():
desc = pick(json.load(open(mf)))
texts[mf] = desc
missing = [c for c in commands if c not in desc] + [s for s in skills if s not in desc]
if missing:
fails += 1
print(f"FAIL: {mf} description omits: {', '.join(missing)}")
else:
print(f"OK: {mf} mentions all {len(commands)} commands + {len(skills)} skills")
claimed = re.search(r'(\d+) skills', desc)
if not claimed:
print(f"OK: {mf} claims no skill count (nothing to drift)")
elif int(claimed.group(1)) != len(skills):
fails += 1
print(f"FAIL: {mf} says '{claimed.group(1)} skills' but {len(skills)} exist")
else:
print(f"OK: {mf} skill count ({len(skills)}) matches reality")
if len(set(texts.values())) != 1:
fails += 1
print("FAIL: the four descriptions are not identical — /release Phase 2 syncs them")
else:
print("OK: all four description homes carry identical text")
sys.exit(1 if fails else 0)
EOF
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.
- 2d ago First seen · 259 lines · 60 tokens per session scan A 41958781b37d
validate is a command published in the GitHub repository PrabhdeepSingh/claude-plugins (3 stars, last pushed 2d ago), licensed MIT. It adds 60 tokens to every session and 3,987 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it A with 1 finding (runs shell commands). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other commands, from other repositories
run-powershell
Execute a PowerShell script on an Atera agent.
log-time
Log work hours on an Atera ticket.
eol-report
EOL/EOS risk report — devices, OS versions, and firmware approaching or past end-of-life/end-of-support, prioritized by criticality.
update-ticket
Update fields on an existing Atera ticket.
alert-triage
Triage open Auvik alerts, rank by severity, and recommend dismissals for known noise.
release
Generate changelog, bump version, and create git tag.