GitHub - joneslloyd/agent-skills

A collection of agent skills. Skills follow the open Agent Skills SKILL.md format and this repository doubles as a Claude Code plugin marketplace.

Skills

counterpart-analysis

Read transcripts, email threads, chat logs, or described situations to infer what each participant actually wants — their underlying interests, constraints, incentives, and likely "angle".

• Quote-level evidence: every motivational claim anchors to a verbatim quote, marker, or structural observation (ignored points, CC changes, effort asymmetry, questions asked).
• Provenance tagging: every input is labeled [OBSERVED] (in the material), [CONTEXT] (supplied background), [INFERRED] (derived, with cited inputs), or [ASSUMED] (unverified prior) — so you can see whether a conclusion rests on fact or assumption, and which assumptions are load-bearing.
• Competing hypotheses: 2–3 ranked hypotheses, each with supporting and disconfirming evidence, and a concrete discriminating test (calibrated question, no-oriented question, or strawman proposal) to run next.
• Calibration guardrails: baseline-first reading, a false-positive table of boring explanations that must be beaten, cross-cultural/ESL downweighting, and hard limits on confidence for thin material.
• Ethics boundary: analysis for understanding and preparation — not manipulation scripts.

Reference files: references/linguistic-markers.md (the "X phrasing hints at Y" catalogue, 16 sections) and references/frameworks.md (interests vs. positions, Black Swans, irrationality diagnostic, incentive mapping, counterpart styles, test design).

Installation

Claude Code (recommended)

/plugin marketplace add joneslloyd/agent-skills
/plugin install counterpart-analysis@agent-skills

Or from the terminal:

claude plugin marketplace add joneslloyd/agent-skills
claude plugin install counterpart-analysis@agent-skills

Update later with /plugin update (or /plugin marketplace update agent-skills).

OpenCode

OpenCode has native SKILL.md support and — usefully — already reads Claude Code's own skill directories, so no path translation is needed:

git clone https://github.com/joneslloyd/agent-skills.git
cp -r agent-skills/plugins/counterpart-analysis/skills/counterpart-analysis ~/.claude/skills/

For a project-local install, copy into .opencode/skills/, .claude/skills/, or .agents/skills/ at your repo root instead (OpenCode checks all three, walking up to the git root). Globally, ~/.config/opencode/skills/, ~/.claude/skills/, and ~/.agents/skills/ are all read. OpenCode has no repo-based marketplace/install command today — see Updating below for how to pull new versions of the skill by hand.

Manual (Codex and other agentskills.io-compatible agents)

Copy the skill folder into your agent's skills directory:

git clone https://github.com/joneslloyd/agent-skills.git
cp -r agent-skills/plugins/counterpart-analysis/skills/counterpart-analysis ~/.claude/skills/

For a repo-local install, use .claude/skills/ in your project instead of ~/.claude/skills/.

Claude.ai / Claude apps

Download the packaged .skill file from Releases and upload it in a conversation — the file card shows a Save skill button (requires an org/plan that allows skill creation), or add it via Settings → Capabilities → Skills.

Updating

The OpenCode and manual-copy paths above have no auto-update mechanism — re-run this whenever you want the latest version:

cd agent-skills && git pull
cp -r plugins/counterpart-analysis/skills/counterpart-analysis ~/.claude/skills/   # or your chosen skills dir

(Claude Code plugin installs update via /plugin update instead — see above. Claude.ai .skill uploads must be re-downloaded and re-uploaded by hand.)

Repository structure

.
├── .claude-plugin/
│   └── marketplace.json          # marketplace catalog
├── plugins/
│   └── counterpart-analysis/
│       ├── .claude-plugin/
│       │   └── plugin.json       # plugin manifest (bump version on release)
│       └── skills/
│           └── counterpart-analysis/
│               ├── SKILL.md
│               └── references/
│                   ├── linguistic-markers.md
│                   └── frameworks.md
├── LICENSE
└── README.md

Releasing

1. Edit the skill under plugins/counterpart-analysis/skills/counterpart-analysis/.
2. Bump version in plugins/counterpart-analysis/.claude-plugin/plugin.json — users only receive updates when this string changes.
3. Validate: claude plugin validate .
4. Commit, push, and (optionally) attach a freshly packaged .skill file to a GitHub Release for claude.ai users.

License

MIT — see LICENSE.