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/ccoalm/ccl-skillsnpx agentmods add commands/ccoalm/ccl-skills/ccl-verify-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/commands/ccoalm/ccl-skills/ccl-verify-skills)<a href="https://agentmods.dev/commands/ccoalm/ccl-skills/ccl-verify-skills"><img src="https://agentmods.dev/badge/commands/ccoalm/ccl-skills/ccl-verify-skills/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/commands/ccoalm/ccl-skills/ccl-verify-skills"><img src="https://agentmods.dev/badge/commands/ccoalm/ccl-skills/ccl-verify-skills.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 | $0.00009 | $0.00792 |
| Opus 5 | $0.00005 | $0.00396 |
| Sonnet 5 | $0.00002 | $0.00158 |
| Haiku 4.5 | $0.00001 | $0.00079 |
Grade C, and why
ccl-verify-skills scanned grade C with 2 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 12d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Recursive force deletehighDestructive command
rm -rf with a variable or a broad path is one typo away from removing the wrong tree.
rm -rf "$tmp_home" "$skills_json" Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
import urllib.parse What it actually says
Verify the CCL skills repository and OpenCode discovery state.
Run these checks from the repository root:
bash -n scripts/install.sh && bash -n scripts/install-opencode.sh
python3 -m json.tool opencode.json >/dev/null
git diff --check
bash skills/skill-extraction-workflow/scripts/check-ccl-skills.sh .
tmp_home=$(mktemp -d)
skills_json=$(mktemp)
real_home=${HOME:-}
HOME="$tmp_home" OPENCODE_DISABLE_EXTERNAL_SKILLS=1 opencode debug skill > "$skills_json"
HOME="$tmp_home" REAL_HOME="$real_home" SKILLS_JSON="$skills_json" python3 - <<'PY'
import json
import os
import urllib.parse
from pathlib import Path
def parse_frontmatter_name(path):
lines = path.read_text(encoding='utf-8').splitlines()
if not lines or lines[0].strip() != '---':
return None
for line in lines[1:]:
if line.strip() == '---':
return None
if line.startswith('name:'):
return line.split(':', 1)[1].strip().strip('"\'') or None
return None
def location_path(value):
text = str(value or '')
if text.startswith('file://'):
return Path(urllib.parse.unquote(urllib.parse.urlparse(text).path)).resolve()
return Path(text).resolve() if text else None
repo = Path.cwd().resolve()
skills_root = repo / 'skills'
expected = {}
for skill_file in sorted(skills_root.glob('*/SKILL.md')):
dirname = skill_file.parent.name
fm_name = parse_frontmatter_name(skill_file)
if fm_name and fm_name != dirname:
raise AssertionError(f'frontmatter name mismatch: {skill_file}: name={fm_name!r} dir={dirname!r}')
expected[dirname] = skill_file.resolve()
skills = json.load(open(os.environ['SKILLS_JSON'], encoding='utf-8'))
actual = {}
for item in skills:
loc = location_path(item.get('location'))
if not loc:
continue
try:
loc.relative_to(skills_root.resolve())
except ValueError:
continue
name = item.get('name') or (loc.parent.name if loc.name == 'SKILL.md' else None)
if name:
actual[str(name)] = loc
expected_names = set(expected)
actual_names = set(actual)
missing = sorted(expected_names - actual_names)
extra = sorted(actual_names - expected_names)
print('expected_local_skill_count=', len(expected_names))
print('actual_local_skill_count=', len(actual_names))
print('missing_local_skills=', ','.join(missing) or '<none>')
print('extra_local_skills=', ','.join(extra) or '<none>')
if missing or extra:
raise AssertionError({'missing': missing, 'extra': extra, 'actual': sorted(actual_names), 'expected': sorted(expected_names)})
real_home = os.environ.get('REAL_HOME')
if real_home:
global_root = Path(real_home).expanduser() / '.config' / 'opencode' / 'skills'
duplicates = sorted(name for name in expected_names if (global_root / name / 'SKILL.md').exists())
if duplicates:
print('warning_global_skill_snapshots=', ','.join(duplicates))
print('warning_global_skill_snapshots_note=these are independent installed snapshots; repo development should trust this isolated local verify. To refresh global OpenCode skills, run install/update and restart OpenCode.')
PY
rm -rf "$tmp_home" "$skills_json"
Report pass/fail honestly. Do not claim OpenCode support is verified if any command fails.
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 · 91 lines · 9 tokens per session scan C 12fe2a2fbf4a
ccl-verify-skills is a command published in the GitHub repository ccoalm/ccl-skills (6 stars, last pushed today), licensed Apache-2.0. It adds 9 tokens to every session and 792 once invoked, about $0.0000 per session on Opus 5. A static security scan graded it C with 2 findings (recursive force delete, makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other commands, from other repositories
learn
Extract patterns and learnings from current session.
workflow-classify
Classify task size (S/M/L/XL) and recommend the appropriate workflow pipeline.
security-review
A security review procedure based on CWE, a list of common software weaknesses, and STRIDE, a method for identifying threats to a system.
atomic-plan
Write a design doc (concepts, business rules, approaches) and a checkpoint-table spec (contract) for non-trivial work; inline spec only for trivial. Gauges triviality; loops spec authoring with subagents. Human-facing artifact, Mermaid diagrams allowed.
export-closedloop-learnings
Exports pending ClosedLoop learnings to global location with deduplication.
review-branch
Review the current branch's diff against base by dispatching atomic-reviewer. No orchestration loop, no spec required — pre-flight before /commit pr or /commit merge.