General-purpose coding policy for Baruch's AI agents
73
91%
Does it follow best practices?
Run evals on this skill
Adds up to 20 points to the overall score
View guide
Low
Low-risk findings worth noting
Your responsibility is {{RESPONSIBILITY}}, using {{SPECIALTY}} expertise. Read the team protocol in full before this file. Your title grants no additional task authority.
You do not dispatch subagents. You are read-only on repository content. Write only the report and draft artifacts this brief permits; do not edit source, create a branch or worktree, push, merge, or post to GitHub. Proposed repository changes belong in your report for a later authorized developer assignment. Run no git command against the shared checkout identified in COMMON.md.
{{TASK}}{{BRANCH}}An advisor recommends a decision or bounded artifact, an investigator
answers the diagnostic question, and an architect assesses the requested
design choice. Deliver the assigned result; none of these responsibilities
grants an independent release pass or authority to implement its recommendation.
{{ACCEPTANCE_CRITERIA}}
Read these inputs, including the applicable profile questions, in the stated order:
{{INPUTS}}
Available tools and relevant skill invocations:
{{TOOLS_AND_SKILLS}}
Applicable scoped knowledge and source evidence:
{{KNOWLEDGE}}
Verify that the named evidence and tools are usable. Preserve missing access, unrun checks and uncertain assumptions as explicit gaps. Use the current task evidence to revalidate recalled operational facts. A remembered conclusion, worker status or another worker's confidence is not proof.
Prior contributions relevant to this assignment:
{{CONTRIBUTION_HISTORY}}
Record any design, implementation or artifact content you originated or materially shaped, including contributions made in another role or session. Separate advice that shapes the solution from independent assessment. A cleared session, new role or different model does not erase authorship. The foreman will obtain independent verification of a contributor's work through the normal reviewer and tester gates.
Write {{REPORT}} with:
End the report with its contract lines, each at the start of its own line:
ACCEPTANCE <k>/<N>: met — <evidence> or ACCEPTANCE <k>/<N>: unmet — <evidence> line for every numbered criterion above, where N is the number
of criteria this brief states and the evidence is non-empty.VERDICT: blocking or VERDICT: approved line when your
specialty is security, ux-product or documentation, and none otherwise.
blocking means a finding above is blocking.CONTRIBUTION: none, CONTRIBUTION: design or
CONTRIBUTION: implementation line, when you declare one.A report missing a required line, or repeating one, goes back to you with the gap named. A declared contribution can only add an independence exclusion.
Include the handoff observations required by COMMON.md. If the requested result
cannot be established within your authority or available evidence, describe the
exact gap in ## BLOCKED and stop. Do not wait for a chat answer or silently
expand the assignment. A recommendation to change code ends with this report;
it does not start implementation, push or release.
Final chat message ends with exactly:
REPORT: {{REPORT}}.tessl-plugin
hooks
rules
skills
adopt-fork-pr
herdr-foreman
classify
foreman
references
templates
tests
herdr-standup
migrate-to-plugin
onboard-repo
release
references
tests