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/martybonacci/specswarm/verifygit clone --depth 1 https://github.com/MartyBonacci/specswarmWhat 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 | $0.00068 | $0.04785 |
| Opus 5 | $0.00034 | $0.02393 |
| Sonnet 5 | $0.00014 | $0.00957 |
| Haiku 4.5 | $0.00007 | $0.00479 |
Grade A, and why
verify scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 3d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 401 lines — stays where its author put it; the contents beside it link to each section on GitHub.
SpecSwarm Spec-Mentor Verification
Runs adversarial verification on a completed task: dispatches a fresh spec-mentor subagent (no carried context, no anchoring bias) to compare what the spec says against what the code does, returns a structured verdict, and updates .specswarm/verify-queue/.
This is the architectural piece that replaces the dual mentor↔builder session pattern. Instead of a long-running mentor session catching drift, a fresh subagent fires once per completed task — better adversarial value, lower context cost, fully automatable.
Phase 1: Resolve target task(s)
PLUGIN_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")/.." && pwd)"
# shellcheck disable=SC1091
source "${PLUGIN_DIR}/lib/verify/queue.sh"
# shellcheck disable=SC1091
source "${PLUGIN_DIR}/lib/verify/task-context.sh"
# shellcheck disable=SC1091
source "${PLUGIN_DIR}/lib/references-loader.sh"
# shellcheck disable=SC1091
[ -f "${PLUGIN_DIR}/lib/notify.sh" ] && source "${PLUGIN_DIR}/lib/notify.sh"
# Parse arguments
TASK_ID=""
PROCESS_ALL=false
FEATURE_OVERRIDE=""
QUEUE_MODE=false
SIGHTED_MODE=false
while [ $# -gt 0 ]; do
case "$1" in
--all|--drain) PROCESS_ALL=true; shift ;;
--queue) QUEUE_MODE=true; shift ;;
--sighted) SIGHTED_MODE=true; shift ;;
--feature) FEATURE_OVERRIDE="$2"; shift 2 ;;
-h|--help)
cat <<EOF
Usage: /ss:verify [TASK_ID] [options]
Adversarial spec-vs-implementation verification for completed task(s).
Options:
--all Process every pending task in the queue
--drain Alias for --all (drains the whole queue)
--feature NUM Override auto-detected feature number
--queue Show queue status (pending / verified / flagged / needs-sighted)
--sighted TASK_ID Human sighted sign-off walkthrough for a NEEDS-SIGHTED task
Examples:
/ss:verify T011
/ss:verify --all
/ss:verify --drain
/ss:verify --queue
/ss:verify --sighted T012
EOF
exit 0
;;
*)
if [ -z "$TASK_ID" ]; then TASK_ID="$1"; fi
shift
;;
esac
done
# Queue overview mode — useful for "how many pending right now?"
if [ "$QUEUE_MODE" = true ]; then
PENDING=$(ss_verify_queue_count pending)
VERIFIED=$(ss_verify_queue_count verified)
FLAGGED=$(ss_verify_queue_count flagged)
NEEDS_SIGHTED=$(ss_verify_queue_count needs-sighted)
echo "📋 SpecSwarm verification queue:"
echo " pending: $PENDING"
echo " verified: $VERIFIED"
echo " flagged: $FLAGGED"
echo " needs-sighted: $NEEDS_SIGHTED"
if [ "$PENDING" -gt 0 ]; then
echo ""
echo "Pending:"
ss_verify_queue_list_pending | sed 's/^/ • /'
fi
if [ "$FLAGGED" -gt 0 ]; then
echo ""
echo "Flagged (need review):"
find "$(ss_verify_queue_dir)" -maxdepth 1 -type f -name '*.flagged' \
-exec basename {} \; 2>/dev/null | sed 's/\.flagged$//' | sed 's/^/ ⚠️ /'
fi
exit 0
fi
# Resolve target task list
TARGETS=()
if [ "$PROCESS_ALL" = true ]; then
while IFS= read -r t; do
[ -n "$t" ] && TARGETS+=("$t")
done < <(ss_verify_queue_list_pending)
elif [ -n "$TASK_ID" ]; then
TARGETS=("$TASK_ID")
else
# Default: oldest pending
OLDEST=$(ss_verify_queue_list_pending | head -n1)
[ -n "$OLDEST" ] && TARGETS=("$OLDEST")
fi
if [ "${#TARGETS[@]}" -eq 0 ]; then
echo "✅ No pending verifications. Queue is empty."
exit 0
fi
echo "🔍 Verification targets: ${TARGETS[*]}"
echo ""
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.
- 3d ago First seen · 401 lines · 68 tokens per session scan A a95318e4684e
verify is a command published in the GitHub repository MartyBonacci/specswarm (65 stars, last pushed 1mo ago), licensed MIT. It adds 68 tokens to every session and 4,785 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 2026-08-30.
Other commands, from other repositories
git
Git operations with intelligent commit messages and workflow optimization.
checklist
Generate a custom checklist for the current feature based on user requirements.
clarify
Identify underspecified areas in the current feature spec by asking up to 5 highly targeted clarification questions and encoding answers back into the spec.
specify
Create or update the feature specification from a natural language feature description.
analyze
Perform a non-destructive cross-artifact consistency and quality analysis across spec.md, plan.md, and tasks.md after task generation.
converge
Assess the current codebase against the feature's spec, plan, and tasks, then append any remaining unbuilt work as new tasks to tasks.md so implement can complete it.