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/anshss/shiploopnpx agentmods add commands/anshss/shiploop/validatedWrote 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/anshss/shiploop/validated)<a href="https://agentmods.dev/commands/anshss/shiploop/validated"><img src="https://agentmods.dev/badge/commands/anshss/shiploop/validated/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/anshss/shiploop/validated"><img src="https://agentmods.dev/badge/commands/anshss/shiploop/validated.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 | $0.00025 | $0.01017 |
| Opus 5.5 | $0.00010 | $0.00407 |
| Sonnet 5.5 | $0.00005 | $0.00203 |
| Haiku 4.5 | $0.00003 | $0.00102 |
Grade A, and why
validated 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 17d 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 — 69 lines — stays where its author put it; the contents beside it link to each section on GitHub.
/shiploop:validated
Defer to the workspace-local copy. If the current workspace has installed its own
.claude/commands/validated.md(every scaffolded shiploop workspace at 1.19.2+ does), follow THAT copy: it is pinned to the workspace's harness version and itsscripts/govern/validation-record.sh. This global copy is the fallback for a workspace that predates the extracted sink writer; the playbook below is identical in shape, but always run the workspace's own script, never a hub path.
Record a ticket as validated: write the durable, git-tracked evidence summary at
.claude/shiploop/validation/ticket-<N>-<slug>.md so this session's proof survives after the ticket
block is gone. $ARGUMENTS is the ticket number <N> (optionally followed by free-text context,
e.g. /validated 94 the restore round-tripped clean).
Load-bearing division of labor: you gather and state the evidence; the write itself goes through
scripts/govern/validation-record.sh, the SAME script the governor's resolve path calls on a
passing autonomous validation. This command is the interactive twin: before it existed, a
ticket you validated BY HAND in a live session recorded nothing durable, since only a
governor-dispatched resolve ever populated the committed sink founder-os context cites as proof.
Before calling the script: is what you have actually evidence?
Reading the source and concluding it looks correct is NOT evidence. Evidence is empirical: a
real id (deployment id, PR number, request id), real command output, a screenshot, a PASS/FAIL
verdict from something you actually ran or observed this session (a live deploy, a gstack browse
walkthrough, a test suite's real output, a database query's actual rows). If all you have is "I read
the code and the logic is right," stop: that is not validation, and recording it as such would
misrepresent what happened. Go run the thing first, or leave the ticket unvalidated.
What to do
- Parse
<N>from$ARGUMENTS(leading integer). Refuse if it isn't a number. - Find the ticket's title. If the ticket's
## #Nheading still exists inqueue/tickets.md(ortickets.mdat the workspace root, check both), read the title from that heading. If the block is already gone (you resolved and deleted it earlier this session), use the title you already know from that work, never invent one. - Collect the PR(s), if any, that this validation rode in on, as
<repo>#<number>(e.g.alpha#142). Zero or more. - State the evidence in your own words: the empirical thing from the section above. Keep it
to what you actually observed: ids, output, verdict. Multi-line evidence is fine; write it to a
scratch file and pass
--evidence-fileinstead of trying to cram it into one shell argument. - Call the script:
Repeatbash scripts/govern/validation-record.sh \ --ticket <N> --title "<title>" \ --evidence "<what you actually observed, PASS/FAIL and all>" \ --pr <repo>#<number> \ --source "interactive session" --gating self--prfor each PR. Omit it entirely when there is none yet.
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.
- 17d ago Changed · +4 tokens per session 930d181c0eeb
- 20d ago First seen · 69 lines · 21 tokens per session scan A 3745a535b330
validated is a command published in the GitHub repository anshss/shiploop (7 stars, last pushed 6d ago), licensed Apache-2.0. It adds 25 tokens to every session and 1,017 once invoked, about $0.0001 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 2026-09-10.
Other commands, from other repositories
tasks
Add to or manage the project task list.
brainstorm
Brainstorm a feature idea into a brief and task.
status
Check team status — what happened, what needs you, what's next.
backlog
The project layer: milestones, epics, and user stories with refusing controllers — spec-seeded, sprint-ready.
status
The state of play, computed fresh: branch, dirty files, the active sprint, open work, index freshness.
camp-init
Create a new camp with the standard directory structure. A camp was previously called a campaign; camp init is the same command either way.