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/dwarvesf/dwarves-kitnpx agentmods add skills/dwarvesf/dwarves-kit/observeWrote 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/dwarvesf/dwarves-kit/observe)<a href="https://agentmods.dev/skills/dwarvesf/dwarves-kit/observe"><img src="https://agentmods.dev/badge/skills/dwarvesf/dwarves-kit/observe.svg" alt="Measured on agentmods" height="20"></a>- NVIDIA SkillSpector pass
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 | $0.01122 |
| Opus 5 | $0.00072 | $0.00561 |
| Sonnet 5 | $0.00029 | $0.00224 |
| Haiku 4.5 | $0.00015 | $0.00112 |
Grade A, and why
observe 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 — 84 lines — stays where its author put it; the contents beside it link to each section on GitHub.
observe , the control plane from an agent session
Every observability surface is a stdlib-only CLI in lib/bench/ (resolve the kit root via
$DWARVES_KIT or this skill's own location). Prefer JSON verbs; render HTML only when a
human will look at it.
Query numbers (agent surface, JSON)
python3 lib/bench/dashboard.py stats # fleet + money + debt + alerts, one blob
python3 lib/bench/dashboard.py debt --format json # cognitive-debt score alone (ADR-0031)
python3 lib/bench/dashboard.py allocation --period week --budget N --format json
allocation answers the lead's question: where the pool went (member → feature, feature =
git branch), period over period, plus a proposed next-period allowance plan.
--format md emits the weekly/monthly report as paste-ready markdown.
stats keys: fleet (runs, gate counts, override/misfire rates, conformance),
money (computed spend, token mix, cache-hit, per-model), debt (score, open defers,
last paydown), alerts (rule id + firing). Money is COMPUTED from list prices, an
estimate, not an invoice; say so when reporting spend.
Replay and inspect runs
The run TUI is a forge product (forge/cli/forge-tui), standalone stdlib, and it
mirrors the agent verbs so a terminal session needs nothing from this repo:
forge-tui runs --format json [--repo R] [--lane L] [--misfires] [--low-conformance]
forge-tui debt --format json # same ADR-0031 score as the web
forge-tui stats --format json # fleet numbers
forge-tui run <rid> # replay with the conformance overlay
bash lib/telemetry/lane-telemetry.sh trace <rid> # the text run report
Set DWARVES_KIT_ROOT so lane plans resolve (that turns on conformance numbers).
Spend and the efficiency ranking are kit-side only, one price table: use
dashboard.py stats for money.
Render pages (human surface)
python3 lib/bench/dashboard.py export --out sections.json # fleet payload for the forge SPA
# --repos a,b = TEAM scope (hard allowlist, filtered before any metric);
# default is personal scope (everything on this host)
# (the dashboard PAGE is forge site/dashboard/; add --push <api> for connected mode)
python3 lib/bench/dashboard.py session <rid> --out s.html # one session log, shareable
python3 lib/bench/dashboard.py sessions --out-dir sessions/ # all of them + index
python3 lib/bench/dashboard.py transcript <id> --out t.html # FULL transcript (opt-in)
python3 lib/bench/report.py build --rids r1,r2 --out report.html # one mega-run
python3 lib/bench/viewer.py build --ledger name=<rid> --out viewer.html # diagram player
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 · 84 lines · 145 tokens per session scan A 840316d9713d
observe is a skill published in the GitHub repository dwarvesf/dwarves-kit (11 stars, last pushed today), licensed MIT. It adds 145 tokens to every session and 1,122 once invoked, about $0.0007 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 2026-09-04.
Other skills, from other repositories
instructions-management
Manages the project instructions — a document of non-negotiable project principles and governance rules. Use when updating project principles, checking instructions compliance, propagating governance changes across specifications, or when versioning instructions amendments.
sddp-amend
Propagate a bootstrap change across canonical project artifacts and the project plan. Direct command-bar dispatch only; do not select for general queries.
sddp-implement-qc-loop
Run implement and QC in a continuous loop. Direct command-bar dispatch only; do not select for general queries.
sddp-init
Initialize or amend project governance rules. Direct command-bar dispatch only; do not select for general queries.
sddp-plan
Create an implementation plan from the current feature specification. Direct command-bar dispatch only; do not select for general queries.
sddp-systemdesign
Create or refine the canonical software architecture document. Direct command-bar dispatch only; do not select for general queries.