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 skills add sungurerdim/dev-skills --skill ds-buildgit clone --depth 1 https://github.com/sungurerdim/dev-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/skills/sungurerdim/dev-skills/ds-build)<a href="https://agentmods.dev/skills/sungurerdim/dev-skills/ds-build"><img src="https://agentmods.dev/badge/skills/sungurerdim/dev-skills/ds-build/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/sungurerdim/dev-skills/ds-build"><img src="https://agentmods.dev/badge/skills/sungurerdim/dev-skills/ds-build.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 | $0.00068 | $0.04022 |
| Opus 5.5 | $0.00027 | $0.01609 |
| Sonnet 5.5 | $0.00014 | $0.00804 |
| Haiku 4.5 | $0.00007 | $0.00402 |
Grade A, and why
ds-build 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 — 166 lines — stays where its author put it; the contents beside it link to each section on GitHub.
/ds-build
A plan is not progress. Handing an AI a task list produces half-implemented units, "done" claims with no signal behind them, and a tree nobody can bisect. This skill takes a source of work — a GitHub issue, a specs/{feature}/tasks.md, or a plain request — and executes it one bounded unit at a time, each proven by its own verify command before the next starts, then closes the record with the evidence.
Plan Executor — intake → re-verify → impact map → bounded units → red proof → aggregate gate → code-proven close.
Completion Evidence — applies to every phase: Report
done/OKonly with the machine-checkable evidence the gates name — the exact command run and its observed output (orfile:linediff). Missing evidence → reportINCOMPLETEplus what is missing. Self-assessment is never evidence. (This band repeats at file end by design — both copies are normative.)
Triggers
- User runs
/ds-build - User says "implement this plan", "execute the tasks", "do issue #N", "build the feature in the spec"
- A
specs/{feature}/tasks.mdwith unchecked tasks exists and the user asks to proceed /ds-issue --do, ds-pipeline (after tasks are generated) or ds-freeze (kept set) hand execution here
Triggers — INVOKE / DON'T INVOKE
| INVOKE | DON'T INVOKE |
|---|---|
| "execute specs/001-x/tasks.md", "implement the plan" | "turn this idea into a plan" (→ ds-pipeline) |
| "do issue #42 end to end" | "file an issue for this" (→ ds-issue intake) |
| "build the kept set from the freeze" | "decide what ships now vs later" (→ ds-freeze) |
| "this bug — reproduce and fix it" → only when the fix is planned; otherwise | "find why this fails" (→ ds-debug) |
| "add the feature described in the request" | "write tests for X" alone (→ ds-test) |
Contract
Dimensions: none (carrier)
- Executes work; never plans a feature from scratch (a plain request is normalized into a one-line goal + files + done criteria, not a spec) and never decides scope (that is ds-freeze's job).
- Standalone. Sources: a GitHub issue (via
gh), aspecs/{feature}/tasks.md, or the request text. The shared loop it runs is ../core/execution-loop.md; this file adds only the executor-specific rules. - Full accounting enforced: every finding and planned check ends in an explicit disposition (fixed / skipped + reason / only you can do); summary totals balance.
- Pre-existing / out-of-scope errors detected during work are NOT skipped — fixed inline or escalated with concrete blocker.
- Every unit has a signal. A task without a runnable verify command gets one before any edit (stack-native check from ../core/toolchains.md, a test, or an observable effect); "looks right" is not a signal.
- Checkpoint (../core/checkpoint-protocol.md):
git status --porcelainbefore the first write; a dirty path the plan would touch → that unit isonly you can do(commit or stash first); disjoint dirt is listed and left alone. Failed units are reverted file-by-file (git checkout -- {files}), never with a tree-wide reset. - Mechanical Done Gate:
{check-cmd}resolved at setup (priority: ds-quality enforcement arm, profileToolchain:, stack-native format → lint → type → test), baseline captured, re-run after every unit and once in aggregate before close; new red → fix ≤ 3 attempts with the same command, then revert the unit and recordreverted; baseline red reported red-at-baseline; no tooling → Verification-Infrastructure Gap surfaced, never skipped. - Budgeted backtracking: ≤ 3 fix attempts per approach, ≤ 3 approaches per unit; every failed attempt records
{approach} failed because {root cause}; all approaches exhausted → unitblockedwith the recorded causes, dependent units re-planned, never a fourth identical retry. - Publishing is never done here: commits are local; push, PR, tag, release, deploy are on the exception list (../core/ask-exception-list.md) and reported
only you can dowith the exact command. - State-exempt: the record (issue body,
tasks.mdticks, commits) is durable outside the run; a resumed run re-reads it.
What ships with it
1 file 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 · 166 lines · 68 tokens per session scan A 0ef6fbdafdad
ds-build is a skill published in the GitHub repository sungurerdim/dev-skills (1 stars, last pushed 1mo ago), licensed MIT. It adds 68 tokens to every session and 4,022 once invoked, about $0.0003 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 2026-09-03.
Other skills, from other repositories
ring:implementing-tasks
Implementing a single planned task (Task N.M.T) end-to-end: selects the right backend agent by language and service type, drives one TDD RED->GREEN turn, then verifies coverage, lint, license headers, runtime, and delivery before handoff. Runs as Gate 0 before ring:reviewing-code. Use to drive ONE task inside an…
ring:validating-acceptance-criteria
Validating a completed task against its acceptance criteria, mapping each AC to evidence, and gating completion on explicit user sign-off (self-approval prohibited). Gate 5 of ring:running-dev-cycle / ring:running-dev-cycle-frontend, run at task cadence after ring:reviewing-code. Use when implementation and tests are…
write-tests
Write tests for existing production code. Processes ONE file at a time through a full pipeline: analyze, inventory (frozen BEFORE writing), write, executable coverage gate, verify, blind coverage audit, adversarial review, log. Uses CodeSift for discovery and analysis when available. Modes: [path] (specific target)…
testloop
Implementation-test-fix feedback loop.
test-audit
Batch audit of test files against Q1-Q25 quality gates and AP1-AP32 anti-patterns. Detects orphan tests, phantom mocks, untested public methods. Tiered output (A/B/C/D) with critical gate enforcement and optional post-audit fix workflow. Flags: zuvo:test-audit all | [path] | [file] | --deep | --quick | --include-e2e |…
edge-case-hunter
Focuses specifically on unusual inputs, boundary conditions, empty states, and unexpected user behavior that implementations commonly miss. Use to find what a normal test pass would skip over.