Upgraded overnight and our workers now leak connections after roughly six hours. Rolling back fixed it.
ReproductionRegressions show up on X before they show up in your issue tracker
When a release goes wrong, the first useful reports are usually posts from strangers hitting the same wall. This skill reads X so a debugging session can check whether the problem is yours alone.
- Bundle
- felo-skills/felo-x-search
- Reads
- Posts, replies, profiles
- Comes back as
- Signals with handles
- Use it when
- You are debugging blind
cp -r felo-skills/felo-x-search .dsh/skills/DeepSeek Harness
Open-source agent harness (developer preview)
dsh reads skills straight out of folders you point it at, so a Felo bundle is one copy away from being callable in a session — and one delete away from being gone.
- Loads from
- ~/project/.dsh/skills/ · ~/.agents/skills/
- Install is
- A folder copy, no build step
- Worth knowing
- Only direct children of a root are read
A feed where every post carries a reason to care
Social results are noisy by default. The skill tags each post with what it contributes — a reproduction, a workaround, a confirmation, or nothing at all.
Same here. Setting the pool to close idle sockets is a workaround until it is patched.
WorkaroundCannot reproduce on 3.1, only on the new async path. Might be limited to long-lived processes.
ScopeMaintainer replied in the issue thread: fix is queued for the next patch release.
StatusAsk for the distinction explicitly: "separate confirmed reports from speculation, and tell me how many people each one comes from."
X is a lead, not a measurement
The failure mode of social data is believing it. Used well, it tells you where to look; used badly, it tells you what a loud minority thinks.
Volume is not prevalence
Four posts about a bug do not mean four percent of users hit it. They mean four people were annoyed enough to type. Pair it with your own error rates before you escalate.
Journalists and bots crowd the results
For anything newsworthy, the top results are commentary rather than experience. Ask for replies from accounts that look like practitioners, not publishers.
Absence proves nothing
No posts about your bug can mean nobody hit it, or that the people who did are not posting. Silence is not a clean bill of health.
Mount it like any other bundle, then keep it out of the way
Unlike the other skills, this one is only useful during a specific kind of session. Project-scoped installation keeps it from showing up in unrelated work.
~/project/.dsh/skills/ProjectLoads for this repository only. The right place to read a bundle before you trust it with the whole machine.
~/.agents/skills/UserLoads in every project, and is the same folder other harnesses on this machine look at.
mkdir -p .dsh/skills
cp -r felo-skills/felo-x-search .dsh/skills/List the skills you can call from this project.The version of this that saves an afternoon
Use it while you are already stuck, not as a monitoring chore.
Check X for reports about this version's connection handling. Group them by what the poster actually observed, and mark which ones include a reproduction.What comes backA grouped picture of the failure that tells you whether you are early, late, or alone — plus a workaround you can try while you wait for a fix.
Find out whether the maintainers have responded publicly about this issue, and link the thread.What comes backThe status, straight from the thread, instead of a guess from the changelog.
Close the loop with something authoritative
Social signals tell you where to look. These two tell you what is actually true.
What dsh users ask about X data
Can it read replies to a specific post?
Does it need my X account?
How fresh is fresh?
Every Felo skill for DeepSeek Harness
Five skills written for DeepSeek Harness. They follow the same pattern on this platform, so you can combine them in one session without re-explaining your setup.
One API key. A full agent toolkit.
Start with a skill, then compose the exact models, live data, knowledge, and output tools your dsh workflow needs — the same key every Felo skill already uses.
Models
Use cost-efficient LLM access when your agent needs custom reasoning or generation.
- LLM API
- Chat completions
- Model access
Live information
Give agents current web and social signals instead of relying on stale context.
- Web Search
- X Search
- Web Fetch
Knowledge and output
Ground work in reusable knowledge, then turn it into shareable deliverables.
- LiveDoc
- PPT generation
- Mindmaps
Creative tools
Extend the workflow from research to visual and office-ready results.
- Image generation
- Page tools
- Office outputs