bondie/agent-harness
A ticket team for Claude Code (analyst, estimator, implementer, reviewer) with approval gates; the estimator forecasts hours from similar finished tickets, backtested against baselines, a local model and human estimates.
ARCHETYPE
Pragmatist
A balanced, no-drama setup: some rules, some tools, nothing extreme.
CONTEXT TAX · EVERY TURN
up to ~3.5k tokens
Claude Code loads MCP tools on demand, so the real cost is usually lower.
Moderate · median rig: 2.3k · breakdown
Copy this rig
# review before running: this installs third-party code $ claude mcp add ticket-precedent -- uvx --from git+https://github.com/bondie/ticket-precedent tp serve $ npx degit bondie/agent-harness/agents ./rig-agent-harness/agents $ npx degit bondie/agent-harness/skills ./rig-agent-harness/skills $ npx degit bondie/agent-harness/hooks ./rig-agent-harness/hooks
MCP servers are added to Claude Code at local scope; env vars are shown as YOUR_… placeholders — we never store values. Files are fetched with degit into a separate folder so you can review before merging.
$ claude mcp add ticket-precedent -- uvx --from git+https://github.com/bondie/ticket-precedent tp serve $ npx degit bondie/agent-harness/skills/retro .claude/skills/retro $ npx degit bondie/agent-harness/skills/stats .claude/skills/stats $ npx degit bondie/agent-harness/skills/ticket .claude/skills/ticket
$ curl -fsSL --create-dirs -o .claude/agents/estimator.md https://raw.githubusercontent.com/bondie/agent-harness/main/agents/estimator.md $ curl -fsSL --create-dirs -o .claude/agents/implementer.md https://raw.githubusercontent.com/bondie/agent-harness/main/agents/implementer.md $ curl -fsSL --create-dirs -o .claude/agents/jira-analyst.md https://raw.githubusercontent.com/bondie/agent-harness/main/agents/jira-analyst.md $ curl -fsSL --create-dirs -o .claude/agents/reviewer-qa.md https://raw.githubusercontent.com/bondie/agent-harness/main/agents/reviewer-qa.md
This rig commits no guardrails. Here is the community baseline instead — the deny/ask rules most often found across all 16.2k rigs:
{
"permissions": {
"deny": [
"Read(./.env)",
"Read(./.env.*)",
"Read(**/.env)",
"Read(~/.ssh/**)",
"Bash(git push --force:*)",
"Read(**/*.pem)",
"Read(.env)",
"Read(~/.aws/**)",
"Read(**/*.key)",
"Read(**/.env.*)",
"Bash(git push --force*)",
"Bash(rm -rf:*)",
"Bash(rm -rf *)",
"Bash(sudo *)",
"Bash(sudo:*)",
"Bash(rm -rf /)",
"Bash(git push -f:*)",
"Read(.env.*)",
"Bash(git reset --hard*)",
"Bash(rm -rf /*)"
],
"ask": [
"Bash(git push:*)",
"Bash(git push *)",
"Bash(gh pr merge *)",
"Bash(git commit:*)",
"Bash(rm *)",
"Bash(gh pr merge:*)",
"Bash(gh pr create *)",
"Bash(git reset --hard:*)",
"Bash(npm publish:*)",
"Bash(git rebase *)"
]
}
} MCP servers (1)
| server | source | est. tokens |
|---|---|---|
| bondie/ticket-precedent · "ticket-precedent" | git | 2.5k |
Skills (3)
Subagents (4)
| estimator model: sonnet | Use this agent to estimate how many hours a ticket will take. With a file-by-file implementation plan it estimates bottom-up, calibrated on the logged time of the person who will do the work; otherwis |
| implementer model: inherit | Use this agent to implement an analysed and approved ticket. Give it the analysis, the approved approach and the acceptance criteria. It changes code and tests, and the plugin's hook runs the project' |
| jira-analyst model: sonnet | Use this agent first when work starts on a tracker ticket. It reads the ticket with its comments, links and attachments, finds precedents and the code involved, and returns an analysis with acceptance |
| reviewer-qa model: sonnet | Use this agent after the implementer finishes, or to review a merge request. It reviews a diff (the uncommitted changes, or a ref range such as origin/dev...origin/feature/X) against the analysis or t |
Similar rigs
HoussemDjeghri/self-review
Claude Code plugin: fresh-context subagents review every change Claude makes, verify their findings and fix what's real, round after round until the review converges. A Stop hook won't let the turn end without it.
Minimalist 216 tok · · Q0
netresearch/retro-skill
LLM-driven session retrospection skill: detects friction in agent sessions and materializes learnings into correct destinations (user-memory, project-rules, skill PRs, checkpoints, harness artefacts)
Minimalist 1.5k tok · · Q0
tsandrini/tensorfiles
tsandrini's fully covariant tensorfiles [maintainer=@tsandrini]
Minimalist 34 tok · · Q0
zmilojevic/hearth
Keep your Claude Code sessions warm. Desktop app for Claude Code, Linux-first (also macOS/Windows): all your sessions in one sidebar, diffs, terminal, cache keep-warm, phone companion. Uses your Claude subscription, no API key.
Pragmatist 14.7k tok · · Q25