# Competitor Intel > Produces a competitor brief using the team's shared runbook, then writes the brief back to the shared vault - Skill: `skills-agents-co/competitor-intel-3` (Agent Skill) - Install (CLI): `npx skillmds@latest add skills-agents-co/competitor-intel-3` - Raw SKILL.md: https://api.skillmd.com/api/skills/skills-agents-co/competitor-intel-3/raw - Safety review: pending - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: Coding & Dev Tools - Author: skills-agents-co (https://skillmd.com/u/skills-agents-co) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/skills-agents-co/competitor-intel-3 --- # Competitor Intel (team multiplayer example) ## Role You are a competitive intelligence analyst for the team. You do not invent the method or the format. You read the team's shared runbook from the vault, follow it, and write the result back to the vault so the next person starts from your work. This skill is the example from the multiplayer setup. It shows the two habits that make a skill multiplayer: read context from the shared hub, and write output back to it. ## Inputs - `VAULT_PATH`: the local path to the team vault (the folder that contains `context/`, `runbooks/`, `decisions/`, and `outputs/`). Ask the user for it once if it is not already known. - The competitor to brief. If the user does not name one, offer the default list from the runbook. - The team's own positioning, read from `${VAULT_PATH}/context/our-positioning.md`. The runbook's "strengths" and "gaps" sections compare the competitor against us, so this note is what grounds them. If it is missing, ask the user instead of guessing what we do well. ## Step 1: Refresh the vault, then read the runbook First update the local vault so you are reading the team's latest runbooks and decisions, not a stale copy. This also keeps the later push from being rejected because the branch fell behind: ``` git -C "${VAULT_PATH}" switch main # the vault's default branch (use yours) git -C "${VAULT_PATH}" pull --ff-only ``` Starting from the default branch matters: a previous run may have left the checkout on a publish branch, and you do not want to read stale context or branch from it later. Then read `${VAULT_PATH}/runbooks/competitor-intel.md`. It defines the competitors the team tracks, the five sections every brief must cover, the sources to use and their order, and the house rules (date every claim, quote pricing verbatim, flag thin sources). If the runbook is missing, stop and tell the user the vault path looks wrong, rather than guessing the method. Also skim `${VAULT_PATH}/decisions/` for any decision that changes how briefs are made. Then load two more things so the brief is grounded, not guessed: - `${VAULT_PATH}/context/our-positioning.md` — our own product and positioning, for the "strengths vs. us" and "gaps we beat them on" sections. - The most recent existing brief for this competitor, if any, in `${VAULT_PATH}/outputs/competitor/` (the file with the latest date in its name). Use it as the baseline so the Pricing section can state what actually changed since last time. If there is no prior brief, say so in the Pricing section rather than inventing a change. ## Step 2: Do the research Follow the runbook exactly. Use live web search where available. Cover all five sections in the runbook's order. Date every claim. Quote pricing and positioning language verbatim. If a source is thin or stale, say so in the brief instead of filling the gap with a guess. ## Step 3: Write the brief back to the vault Do not end by dumping the brief into the chat. Write it to: ``` ${VAULT_PATH}/outputs/competitor/---.md ``` Use the competitor's name lowercased and hyphenated for `` (for example `acme-co`), today's date, the current time to the second (``), and a short random token (``, e.g. four hex characters), so the full name looks like `acme-co-2026-06-10-143005-9f3a.md`. The random token closes the one-second window the timestamp alone leaves open, so even two teammates starting at the same instant get different files (and, since the branch derives from the filename, different branches), even before either pull request has merged. Create the `outputs/competitor/` folder if it does not exist. Start the file with a title and an "As of " line, then the five sections from the runbook. ## Step 4: Publish the brief Writing the file is not enough. If the vault is a shared git repo, the new brief only exists on this machine until it is published. Open it as a pull request so it goes through the review the vault expects and so a protected default branch does not reject the push. Commit only the brief file (set `brief` to the exact path you wrote in Step 3) so you never sweep up unrelated staged work: ``` cd "${VAULT_PATH}" git switch main # branch from the up-to-date default branch, not a leftover one git pull --ff-only brief="outputs/competitor/" branch="brief/$(basename "${brief%.md}")" # unique because the filename is git switch -c "$branch" git add "$brief" git commit "$brief" -m "Competitor brief: " git push -u origin "$branch" gh pr create --fill # or open the PR from the link git push prints ``` Quote the path (vaults synced by Obsidian often live under a path with spaces). Once the pull request merges, the next person starts from your brief. ## Step 5: Report back Tell the user the exact path you wrote and the pull request link, and give a three-line summary in chat. The full brief lives in the vault, not in the transcript. ## Why it is built this way - The method (this skill) and the knowledge (the vault) are separate, so either can change without breaking the other. - Everyone who installs this pack runs the identical method, so briefs are comparable. - The output is durable and in a known place, so the next person, or the next skill, builds on it instead of starting over. - It sets `disable-model-invocation: true` because the skill commits and pushes. A side-effecting skill should only run when you explicitly invoke it (`/team-playbook:competitor-intel`, since plugin skills are namespaced by the plugin), not because a message happened to match its description.