aschuth

4 ms·

1. 1.▲
   Show HN: Skills – Making AI coding tools aware of government standards(anneschuth.nl)

   6 points by aschuth 7mo ago|1 comments
2. 2.▲
   by aschuth 8mo ago

   Fair concern. Pinchwork itself just passes text around (task descriptions and results). No keys are shared. But you're right that a malicious task could ask a worker agent to do something dangerous ("run this script", "c
3. 3.▲
   by aschuth 8mo ago

   Thanks!
4. 4.▲
   by aschuth 8mo ago

   Thanks! Right now pricing is set by the poster — you decide how many credits a task is worth. It's intentionally simple: if your bounty is too low, no one picks it up, so you raise it. Market price discovery through trial and error. Th
5. 5.▲
   by aschuth 8mo ago

   Thanks!
6. 6.▲
   Show HN: Pinchwork – A task marketplace where AI agents hire each other(github.com)

   12 points by aschuth 8mo ago|8 comments
7. 7.▲
   Show HN: Claude Threads – Collaborate on Claude Code via Slack (Or Mattermost)(claude-threads.run)

   1 points by aschuth 8mo ago|0 comments
8. 8.▲
   Show HN: Claude Threads – Collaborate on Claude Code via Slack/Mattermost(claude-threads.run)

   1 points by aschuth 8mo ago|0 comments
9. 9.▲
   by aschuth 9mo ago

   The skill/CLI argument misses what MCP enables for interactive workflows. Sure, Claude can shell out to psql. But MCP lets you build approval gates, audit logs, and multi-step transactions that pause for human input. Claude Code's