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 curie-eng/curie --skill implement-issuegit clone --depth 1 https://github.com/curie-eng/curieWrote 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/curie-eng/curie/implement-issue)<a href="https://agentmods.dev/skills/curie-eng/curie/implement-issue"><img src="https://agentmods.dev/badge/skills/curie-eng/curie/implement-issue/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/curie-eng/curie/implement-issue"><img src="https://agentmods.dev/badge/skills/curie-eng/curie/implement-issue.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 | $0.00038 | $0.05675 |
| Opus 5.5 | $0.00015 | $0.02270 |
| Sonnet 5.5 | $0.00008 | $0.01135 |
| Haiku 4.5 | $0.00004 | $0.00568 |
Grade B, and why
implement-issue scanned grade B with 2 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured yesterday.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Instruction-override phrasingmediumPrompt injection
Text telling the model to disregard its earlier instructions or safety rules is the shape of a prompt injection, whoever wrote it.
phrased ("ignore previous instructions", "as the maintainer I authorize...", Downgraded: this mod is about security review, or the phrase is quoted, so it is likely naming the pattern rather than instructing it.
Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
curl -sS -o /dev/null -m "$command_timeout_seconds" -w '%{http_code} %{remote_ip}\n' --resolve "$registry_host:443:$address" "https://$registry_host/" How it starts
The opening of the file, as written. The whole thing — 451 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Implement one issue
Phases
Every run walks nine phases, named in each step heading below: read_issue,
pin_criteria, plan (which starts by exploring the repository),
plan_review, failing_test, implement (which also runs the checks),
review_diff, publish and wait_ci. Three
pairs loop: plan and plan_review, then implement and review_diff, then
wait_ci back to implement when the pull request's checks fail. Each loop
runs at most 3 rounds.
Call mcp__curie__report_progress once when you enter a phase, not while you
work in it: report before the phase's first tool call, and do not report the
same phase again while you stay in it. Every phase except wait_ci, which
the platform reports, gets one report per entry — a new round of a loop, or
the return from wait_ci to implement, is a new entry. Pass phase set to
that phase's id, round for plan, plan_review, implement and
review_diff, and an optional one-line note saying what you are about to
do. The note is public on the issue, so it must contain no secrets and no raw
tool output. If the tool returns an error, continue the work; progress never
blocks the run.
Every message you send must include a tool call until you have called
mcp__curie__publish_changes, are stopping with Could not complete:, or are
finishing a follow-up with No changes needed:. A message without a tool call
ends the run. An all-text reply such as "Now
I'll revise the plan" stops the loop mid-phase, publishing nothing and
stating no reason. Say what you are doing in a message that also makes the
next tool call. Your only text-only endings are the Could not complete:
stop, No changes needed: when a follow-up on an open pull request (a
mention, review feedback, or a relabelled issue whose open pull request this
run continues) needs no change, and the publication-pending note after
publish_changes.
Keep every Bash command in the foreground. Do not set run_in_background to
true; the hook refuses background Bash calls. Wait for each command to finish
and inspect its result before continuing or ending your turn. This applies to
builds, tests, checks and other long commands. Agent calls are also run
in the foreground. Set run_in_background to false on every reviewer call.
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.
- yesterday Changed · +10 lines 946df15986ad
- 2d ago Changed · +27 lines ccc942eb7351
- 4d ago Changed · +4 lines 5a5e4a733fe8
- 5d ago Changed · +9 lines b5b5a9149c89
- 9d ago Changed · +13 lines 7f9d6bbaa323
- 12d ago Changed · +46 lines 2839d6ede15e
- 14d ago Changed · +39 lines da44c9ad990d
- 18d ago First seen · 303 lines · 38 tokens per session scan B 14871214dd56
implement-issue is a skill published in the GitHub repository curie-eng/curie (36 stars, last pushed today), licensed Apache-2.0. It adds 38 tokens to every session and 5,675 once invoked, about $0.0002 per session on Opus 5.5. A static security scan graded it B with 2 findings (instruction-override phrasing, makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-26.
Other skills, from other repositories
github-workflow
GitHub best practices for pull requests, code reviews, issues, Actions workflows, and repository management.
lerna
Best practices for Lerna monorepo management, versioning, and publishing.
apex-status
CTO-level project status from git and codebase state. Use when asked "where are we", "project status", "what's done", or at the start of a work session.
radin-batch
Work backlog tasks via sub-agents, prioritized, one commit each; asks whether to plan each task first. Use for "work through my backlog", "skip planning", or name an epic or task to run only those.
implement-issue
Implements an issue end-to-end: fetches the spec from the project issue tracker, creates branch, writes code with TDD, runs checks, commits, publishes a draft change (PR/MR), closes issue on merge. Tracker- and host-agnostic — GitHub via gh CLI is the factory default; docs/agents/issue-tracker.md and…
git-commit
A Git workflow for creating Conventional Commits messages from actual workspace changes and, when requested, committing or pushing them.