skills

agent
Security Audit
Warn
Health Warn
  • License — License: MIT
  • Description — Repository has a description
  • Active repo — Last push 0 days ago
  • Low visibility — Only 7 GitHub stars
Code Warn
  • Code scan incomplete — No supported source files were scanned during light audit
Permissions Pass
  • Permissions — No dangerous permissions requested

No AI report is available for this listing yet.

SUMMARY

Personal skills for coding agents.

README.md

Colony

General methods, indexed by the moment they fire. Each skill is a folder with a SKILL.md: the name and description are the trigger spec — the user's moment, not a summary. The body anchors each mechanism to its original source with the key quotation.

Every skill sets disable-model-invocation: true: they run when invoked, not auto-suggested.

Install

No installer required. Copy any skill folder into your harness's skills directory:

cp -r is-this-a-thing ~/.claude/skills/     # Claude Code
cp -r is-this-a-thing ~/.agents/skills/   # agents-compatible harnesses
cp -r is-this-a-thing .devin/skills/      # Devin

Or manage the set with skill-loadout.

The moments

The request isn't clear yet

Skill Use when
read-my-mind a short, vague or corrective message leaves intent and scope unspoken — infer first, ask last
ask-small a question is owed: the cheapest comparison that resolves the stuck preference
play-it-back restate the goal so the user can confirm or correct it before work starts
what-are-we-assuming a plan or judgment rests on unstated premises — including when the request is a solution, not the problem
be-more-ambitious a task reads like a deliverables list — raise altitude, depth or range

Look before you build

Skill Use when
is-this-a-thing a method, rule or practice may already exist under another name — find the name and original source
deep-research a consequential question needs field reports and primary evidence, not generic summaries
when-in-rome the codebase already has conventions — copy the house style, not a fresh invention

Design decisions

Skill Use when
functional-programming modeling state so illegal states can't be constructed; deciding where effects live
let-it-crash failure-path contracts: who owns each check, where validation lives, what fails loudly
performance-engineering workload cost modeling, or a claimed improvement that needs proof
open-from-day-one a project, interface or repo that outsiders will ever touch
close-the-design an architecture, spec vocabulary or abstraction is about to become load-bearing — audit the five closures

Delivering to people

Skill Use when
write-the-docs a README, design doc, module doc or task record
for-the-reader turning working materials into an artifact its readers can use
say-it-plain text that is too long, too jargon-dense or too clever for its reader
managing-up reporting to or escalating a decision to a delegator who did not follow the work
show-your-math a number or default needs defending — derivation shown, calibration separated

Trust but verify

Skill Use when
review-the-pr a PR produced faster than anyone can read — machines verify, context checks cover the rest, a person signs the risk
straight-answer a reply that may be defensive, padded or calibrated to please rather than to be right
prove-wrong a conclusion that has only been confirmed — attack it until it breaks or earns credence
slim an artifact that may be mostly removable — prove how much by deleting and re-running

Meta

Skill Use when
skill-creator creating, splitting, renaming or pruning a skill — mechanics in MECHANICS.md

Sources and verification

*/sources/ holds fair-use slices — the quoted passages plus context — so a reader can check each quotation against its original. The full texts are not committed; each citation in SKILL.md is the locator to fetch them.

License

MIT

Reviews (0)

No results found