report-new

A command for starting an analysis report in a separate documentation repository, using information from one or more recorded experiment runs.

In plain words
What is it for?
Use it to create reports for one experiment run, compare runs or branches, document parameter studies, summarize random-seed runs, or record milestones.
Why use it?
It avoids manually copying run references and metrics into a report and helps distinguish single-run, comparison, ablation, seed-sweep, and milestone reports.

Command

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add commands/zhyx12/projtool/report-new
Clone the repo
git clone --depth 1 https://github.com/zhyx12/projtool
Per session 29 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 989 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00029 $0.00989
Opus 5 $0.00015 $0.00495
Sonnet 5 $0.00006 $0.00198
Haiku 4.5 $0.00003 $0.00099

Measured 2d ago against content hash fbde99067a81, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

report-new scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

src/projtool/assets/project_templates/claude/commands/report-new.md · 74 lines

How it starts

The opening of the file, as written. The whole thing — 74 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are running the user's /report-new command.

Steps

  1. Ask the user which run(s) the report covers. Accept either explicit run_ids or "the last run", "all of exp-X", "compare runs A and B".

    • For implicit selectors, call mcp__projtool__list_runs with the appropriate filters and confirm with the user before continuing.
    • If a run is is_dirty=true, mention it — the report frontmatter will get a provenance_warnings field.
  2. Decide the report type and scope:

    • type: single, comparison, ablation, seed_sweep, milestone. If unsure, ask. Default for one run is single; for >1 run on the same branch, ask whether it's an ablation or comparison; for runs spanning multiple branches, the report is a cross-branch summary.
    • scope: branch (default — report belongs to one code-repo branch, filed under that branch's directory in the docs repo) or summary (cross-branch — filed under summary/).
  3. Call mcp__projtool__start_report with experiment_id, run_ids, report_type, and scope. The tool:

    • Loads each run's manifest (via load_manifest).
    • Generates a report frontmatter prefilled with references and metrics_inlined skeletons, plus the cross-repo fields git.branch, git.commit, git.code_repo_url, git.commit_url.
    • Writes the file to the docs repo (NOT to the code repo!). The path is either <docs_repo>/<branch_full>/<date>-<topic_slug>.md (scope=branch) or <docs_repo>/summary/<date>-<topic_slug>.md (scope=summary), where <docs_repo> = <container_root>/docs/ (the independent docs git repo at the container root, alongside code/ and output/). The slug is initially "TBD"; pass a real slug to commit_report later.
  4. Show the user the path to the new file. Open it (with view) so they can see the frontmatter and start drafting the body.

  5. Reference handling:

    • When the body cites a number, also fill the corresponding field in metrics_inlined. If a number is in the body but missing from frontmatter, that's a violation — call it out.
    • If the run group is >8 runs, the tool will have generated a data/<sweep-name>.csv (alongside the report, in the same docs-repo subdirectory) instead of inlining the list; tell the user.
  6. Do not commit yet. The report is a draft. Tell the user to use /report-commit (or mcp__projtool__commit_report) once they're done. commit_report will:

    • Validate that all cited code commits are pushed to the code repo's origin. If not, it errors with a clear instruction to git push first (cross-repo invariant).
    • git add the specific .md and (if externalized) CSV files, then commits and pushes to the docs repo's current branch (resolved via git symbolic-ref --short HEAD, never hardcoded as main) in the docs repo (NOT the code repo).

Read the full file on GitHub · 74 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 2d ago First seen · 74 lines · 29 tokens per session scan A fbde99067a81

Subscribe to this mod's changes

report-new is a command published in the GitHub repository zhyx12/projtool (1 stars, last pushed 19d ago), licensed MIT. It adds 29 tokens to every session and 989 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.